feat(ci): optimize web image build with multi-stage Dockerfile + buildx cache

- staging构建额外打:staging tag并push到Registry
- staging容器加com.centurylinklabs.watchtower.enable=true标签
- 容器用完整Registry路径+:staging tag启动，供Watchtower监控
- 1分钟轮询，自动检测新镜像并重启

.gitea/workflows/test-ssh-secret.yml

@@ -1,69 +0,0 @@
-name: Test SSH Secret
-on:
-  push:
-    branches: [develop]
-    paths:
-      - '.gitea/workflows/test-ssh-secret.yml'
-
-jobs:
-  test-ssh:
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Install SSH client
-        run: |
-          which ssh || (apt-get update && apt-get install -y openssh-client)
-          ssh -V
-
-      - name: Debug environment
-        run: |
-          echo "=== Environment ==="
-          echo "Runner hostname: $(hostname)"
-          echo "Runner IP: $(hostname -i || echo 'unknown')"
-          echo "Current user: $(whoami)"
-          echo "=== Secrets check ==="
-          if [ -n "$STAGING_SSH_HOST" ]; then
-            echo "STAGING_SSH_HOST: [SET] value_length=${#STAGING_SSH_HOST}"
-          else
-            echo "STAGING_SSH_HOST: [EMPTY]"
-          fi
-          if [ -n "$STAGING_SSH_USER" ]; then
-            echo "STAGING_SSH_USER: [SET] value_length=${#STAGING_SSH_USER}"
-          else
-            echo "STAGING_SSH_USER: [EMPTY]"
-          fi
-          if [ -n "$STAGING_SSH_KEY" ]; then
-            echo "STAGING_SSH_KEY: [SET] value_length=${#STAGING_SSH_KEY}"
-          else
-            echo "STAGING_SSH_KEY: [EMPTY]"
-          fi
-        env:
-          STAGING_SSH_HOST: ${{ secrets.STAGING_SSH_HOST }}
-          STAGING_SSH_USER: ${{ secrets.STAGING_SSH_USER }}
-          STAGING_SSH_KEY: ${{ secrets.STAGING_SSH_KEY }}
-
-      - name: Setup SSH key
-        run: |
-          mkdir -p ~/.ssh
-          chmod 700 ~/.ssh
-          echo "$STAGING_SSH_KEY" > ~/.ssh/id_ed25519
-          chmod 600 ~/.ssh/id_ed25519
-          ssh-keygen -y -f ~/.ssh/id_ed25519 > ~/.ssh/id_ed25519.pub 2>/dev/null || echo "No public key generated"
-          echo "=== SSH Key fingerprint ==="
-          ssh-keygen -lf ~/.ssh/id_ed25519 || echo "Key fingerprint failed"
-        env:
-          STAGING_SSH_KEY: ${{ secrets.STAGING_SSH_KEY }}
-
-      - name: Test SSH connection
-        run: |
-          echo "Attempting SSH connection to $STAGING_SSH_HOST..."
-          ssh -i ~/.ssh/id_ed25519 \
-            -o StrictHostKeyChecking=no \
-            -o UserKnownHostsFile=/dev/null \
-            -o ConnectTimeout=10 \
-            -o BatchMode=yes \
-            -v \
-            $STAGING_SSH_USER@$STAGING_SSH_HOST "echo 'SSH_CONNECTION_SUCCESS' && hostname && whoami"
-          echo "=== SSH Test Complete ==="
-        env:
-          STAGING_SSH_HOST: ${{ secrets.STAGING_SSH_HOST }}
-          STAGING_SSH_USER: ${{ secrets.STAGING_SSH_USER }}

alembic/versions/026_add_user_profile_fields.py

@@ -0,0 +1,56 @@
+"""Add user profile fields (name, avatar, updated_at)
+
+Revision ID: 026
+Revises: 025
+Create Date: 2026-07-05
+
+补录用户资料字段。生产数据库已手动添加过这些字段，
+因此 upgrade 做幂等检查，避免在已有字段的库上执行报错。
+"""
+
+import sqlalchemy as sa
+
+from alembic import context, op
+
+revision = "026"
+down_revision = "025"
+branch_labels = None
+depends_on = None
+
+
+def _column_exists(table: str, column: str) -> bool:
+    if context.is_offline_mode():
+        return False
+    conn = op.get_bind()
+    result = conn.execute(
+        sa.text(
+            "SELECT COUNT(*) FROM information_schema.columns " "WHERE table_name = :table AND column_name = :column"
+        ),
+        {"table": table, "column": column},
+    )
+    return result.scalar() > 0
+
+
+def upgrade() -> None:
+    if not _column_exists("users", "name"):
+        op.add_column("users", sa.Column("name", sa.String(100), nullable=True))
+
+    if not _column_exists("users", "avatar"):
+        op.add_column("users", sa.Column("avatar", sa.String(500), nullable=True))
+
+    if not _column_exists("users", "updated_at"):
+        op.add_column(
+            "users",
+            sa.Column(
+                "updated_at",
+                sa.DateTime(),
+                nullable=True,
+                server_default=sa.func.now(),
+            ),
+        )
+
+
+def downgrade() -> None:
+    op.drop_column("users", "updated_at")
+    op.drop_column("users", "avatar")
+    op.drop_column("users", "name")

alembic/versions/027_add_user_ban_fields.py

@@ -0,0 +1,44 @@
+"""Add user ban fields (ban_reason, ban_at)
+
+Revision ID: 027
+Revises: 026
+Create Date: 2026-07-05
+
+补录用户封禁字段。生产数据库已手动添加过这些字段，
+因此 upgrade 做幂等检查。
+"""
+
+import sqlalchemy as sa
+
+from alembic import context, op
+
+revision = "027"
+down_revision = "026"
+branch_labels = None
+depends_on = None
+
+
+def _column_exists(table: str, column: str) -> bool:
+    if context.is_offline_mode():
+        return False
+    conn = op.get_bind()
+    result = conn.execute(
+        sa.text(
+            "SELECT COUNT(*) FROM information_schema.columns " "WHERE table_name = :table AND column_name = :column"
+        ),
+        {"table": table, "column": column},
+    )
+    return result.scalar() > 0
+
+
+def upgrade() -> None:
+    if not _column_exists("users", "ban_reason"):
+        op.add_column("users", sa.Column("ban_reason", sa.Text(), nullable=True))
+
+    if not _column_exists("users", "ban_at"):
+        op.add_column("users", sa.Column("ban_at", sa.DateTime(), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("users", "ban_at")
+    op.drop_column("users", "ban_reason")

alembic/versions/028_add_user_admin_fields.py

@@ -0,0 +1,44 @@
+"""Add user admin fields (admin_status, admin_remarks)
+
+Revision ID: 028
+Revises: 027
+Create Date: 2026-07-05
+
+补录管理员备注字段。生产数据库已手动添加过这些字段，
+因此 upgrade 做幂等检查。
+"""
+
+import sqlalchemy as sa
+
+from alembic import context, op
+
+revision = "028"
+down_revision = "027"
+branch_labels = None
+depends_on = None
+
+
+def _column_exists(table: str, column: str) -> bool:
+    if context.is_offline_mode():
+        return False
+    conn = op.get_bind()
+    result = conn.execute(
+        sa.text(
+            "SELECT COUNT(*) FROM information_schema.columns " "WHERE table_name = :table AND column_name = :column"
+        ),
+        {"table": table, "column": column},
+    )
+    return result.scalar() > 0
+
+
+def upgrade() -> None:
+    if not _column_exists("users", "admin_status"):
+        op.add_column("users", sa.Column("admin_status", sa.String(50), nullable=True))
+
+    if not _column_exists("users", "admin_remarks"):
+        op.add_column("users", sa.Column("admin_remarks", sa.Text(), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("users", "admin_remarks")
+    op.drop_column("users", "admin_status")

alembic/versions/029_add_user_phone.py

@@ -0,0 +1,40 @@
+"""Add user phone field
+
+Revision ID: 029
+Revises: 028
+Create Date: 2026-07-05
+
+补录用户手机号字段。生产数据库已手动添加过该字段，
+因此 upgrade 做幂等检查。
+"""
+
+import sqlalchemy as sa
+
+from alembic import context, op
+
+revision = "029"
+down_revision = "028"
+branch_labels = None
+depends_on = None
+
+
+def _column_exists(table: str, column: str) -> bool:
+    if context.is_offline_mode():
+        return False
+    conn = op.get_bind()
+    result = conn.execute(
+        sa.text(
+            "SELECT COUNT(*) FROM information_schema.columns " "WHERE table_name = :table AND column_name = :column"
+        ),
+        {"table": table, "column": column},
+    )
+    return result.scalar() > 0
+
+
+def upgrade() -> None:
+    if not _column_exists("users", "phone"):
+        op.add_column("users", sa.Column("phone", sa.String(20), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("users", "phone")

alembic/versions/030_add_tags_and_asset_tags.py

@@ -0,0 +1,68 @@
+"""Add tags and asset_tags tables
+
+Revision ID: 030
+Revises: 029
+Create Date: 2026-07-07
+
+新增标签表和素材-标签关联表，支持规范化多对多标签管理。
+"""
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision = "030"
+down_revision = "029"
+branch_labels = None
+depends_on = None
+
+
+def _table_exists(table: str) -> bool:
+    ctx = op.get_context()
+    if ctx.as_sql:
+        return False
+    conn = op.get_bind()
+    result = conn.execute(
+        sa.text("SELECT COUNT(*) FROM information_schema.tables WHERE table_name = :table"),
+        {"table": table},
+    )
+    return (result.scalar() or 0) > 0
+
+
+def upgrade() -> None:
+    if not _table_exists("tags"):
+        op.create_table(
+            "tags",
+            sa.Column("id", sa.String(36), primary_key=True),
+            sa.Column("user_id", sa.String(36), nullable=False),
+            sa.Column("name", sa.String(100), nullable=False),
+            sa.Column(
+                "created_at",
+                sa.DateTime(),
+                nullable=False,
+                server_default=sa.func.now(),
+            ),
+            sa.UniqueConstraint("user_id", "name", name="uq_tags_user_name"),
+        )
+        op.create_index("ix_tags_user_id", "tags", ["user_id"])
+
+    if not _table_exists("asset_tags"):
+        op.create_table(
+            "asset_tags",
+            sa.Column("asset_id", sa.String(36), primary_key=True),
+            sa.Column("tag_id", sa.String(36), primary_key=True),
+            sa.Column(
+                "created_at",
+                sa.DateTime(),
+                nullable=False,
+                server_default=sa.func.now(),
+            ),
+        )
+        op.create_index("ix_asset_tags_tag_id", "asset_tags", ["tag_id"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_asset_tags_tag_id", table_name="asset_tags")
+    op.drop_table("asset_tags")
+    op.drop_index("ix_tags_user_id", table_name="tags")
+    op.drop_table("tags")

alembic/versions/031_add_file_hash_to_assets_and_ing_jobs.py

@@ -0,0 +1,33 @@
+"""Add file_hash to assets and ingest_jobs
+
+Revision ID: 031
+Revises: 030
+Create Date: 2026-07-07
+
+为素材去重检测功能添加 file_hash 字段。
+"""
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision = "031"
+down_revision = "030"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column("assets", sa.Column("file_hash", sa.String(64), nullable=True))
+    op.create_index(op.f("ix_assets_file_hash"), "assets", ["file_hash"])
+
+    op.add_column("ingest_jobs", sa.Column("file_hash", sa.String(64), nullable=True))
+    op.create_index(op.f("ix_ingest_jobs_file_hash"), "ingest_jobs", ["file_hash"])
+
+
+def downgrade() -> None:
+    op.drop_index(op.f("ix_ingest_jobs_file_hash"), table_name="ingest_jobs")
+    op.drop_column("ingest_jobs", "file_hash")
+
+    op.drop_index(op.f("ix_assets_file_hash"), table_name="assets")
+    op.drop_column("assets", "file_hash")

alembic/versions/032_add_asset_select_mode_to_generation_tasks.py

@@ -0,0 +1,28 @@
+"""Add asset_select_mode to generation_tasks
+
+Revision ID: 032
+Revises: 031
+Create Date: 2026-07-07
+
+素材库自动匹配功能：为 generation_tasks 表添加 asset_select_mode 字段。
+"""
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision = "032"
+down_revision = "031"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "generation_tasks",
+        sa.Column("asset_select_mode", sa.String(20), nullable=False, server_default=""),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("generation_tasks", "asset_select_mode")

alembic/versions/033_add_batch_id_to_generation_tasks.py

@@ -0,0 +1,31 @@
+"""Add batch_id to generation_tasks
+
+Revision ID: 033
+Revises: 032
+Create Date: 2026-07-07
+
+视频查重功能：为 generation_tasks 表添加 batch_id 字段，
+用于关联同一次批量生成请求中的多个任务。
+"""
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision = "033"
+down_revision = "032"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "generation_tasks",
+        sa.Column("batch_id", sa.String(32), nullable=False, server_default=""),
+    )
+    op.create_index(op.f("ix_generation_tasks_batch_id"), "generation_tasks", ["batch_id"])
+
+
+def downgrade() -> None:
+    op.drop_index(op.f("ix_generation_tasks_batch_id"), table_name="generation_tasks")
+    op.drop_column("generation_tasks", "batch_id")

apps/api/app/api/router.py

@@ -16,6 +16,7 @@ from app.api.routes.jobs import router as jobs_router
 from app.api.routes.projects import router as projects_router
 from app.api.routes.recipes import router as recipes_router
 from app.api.routes.subscription import router as subscription_router
+from app.api.routes.tags import router as tags_router
 from app.api.routes.task_center import router as task_center_router
 from app.api.routes.templates import router as templates_router
 from app.api.routes.titles import router as titles_router
@@ -38,6 +39,11 @@ api_router.include_router(
     prefix="/projects",
     tags=["Project"],
 )
+api_router.include_router(
+    tags_router,
+    prefix="/tags",
+    tags=["Tag"],
+)
 api_router.include_router(
     task_center_router,
     tags=["TaskCenter"],

apps/api/app/api/routes/asset_libraries.py

@@ -5,6 +5,7 @@ from app.dependencies import get_asset_library_repository, get_project_repositor
 from app.schemas.asset_library import (
     AssetLibraryResponse,
     CreateAssetLibraryRequest,
+    EnsureDefaultLibraryRequest,
     ListAssetLibrariesResponse,
 )
 from fastapi import APIRouter, Depends, HTTPException, Query, status
@@ -15,7 +16,7 @@ from packages.application import (
     GetProjectUseCase,
     ListAssetLibrariesUseCase,
 )
-from packages.domain import AssetLibraryKind
+from packages.domain import AssetLibrary, AssetLibraryKind
 
 router = APIRouter()
 
@@ -43,6 +44,7 @@ def _to_asset_library_response(item) -> AssetLibraryResponse:
 @router.get("", response_model=ListAssetLibrariesResponse)
 def list_asset_libraries(
     project_id: str | None = Query(None),
+    kind: str | None = Query(None, pattern="^(video|voice|image)$"),
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     asset_library_repository: Any = Depends(get_asset_library_repository),
     project_repository: Any = Depends(get_project_repository),
@@ -66,6 +68,11 @@ def list_asset_libraries(
             all_items.extend(use_case.execute(proj.id))
         items = all_items
 
+    # 按 kind 过滤（可选）
+    if kind:
+        kind_enum = AssetLibraryKind(kind)
+        items = [item for item in items if item.kind == kind_enum]
+
     return ListAssetLibrariesResponse(items=[_to_asset_library_response(item) for item in items])
 
 
@@ -90,3 +97,53 @@ def create_asset_library(
         )
     )
     return _to_asset_library_response(item)
+
+
+# 默认素材库名称映射
+_DEFAULT_LIBRARY_NAMES = {
+    "video": "视频素材库",
+    "voice": "配音素材库",
+    "image": "图片素材库",
+}
+
+
+@router.post("/ensure-default", response_model=AssetLibraryResponse)
+def ensure_default_library(
+    request: EnsureDefaultLibraryRequest,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_library_repository: Any = Depends(get_asset_library_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> AssetLibraryResponse:
+    """确保项目下指定 kind 的默认素材库存在，已存在则直接返回，不存在则自动创建。"""
+    project = project_repository.find_by_id(request.project_id)
+    if project is None:
+        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="Project not found")
+    if not project.can_access(authenticated_user.user.id):
+        raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail="Access denied to project")
+
+    kind = AssetLibraryKind(request.kind)
+
+    # 查找该项目下同 kind 的素材库，返回第一个
+    existing = asset_library_repository.find_by_project(request.project_id)
+    for lib in existing:
+        if lib.kind == kind:
+            return _to_asset_library_response(lib)
+
+    # 不存在 → 自动创建
+    import uuid
+    from datetime import datetime, timezone
+
+    now = datetime.now(timezone.utc)
+    default_name = _DEFAULT_LIBRARY_NAMES.get(request.kind, f"{request.kind}素材库")
+    library = AssetLibrary(
+        id=str(uuid.uuid4()),
+        project_id=request.project_id,
+        name=default_name,
+        kind=kind,
+        asset_count=0,
+        total_size=0,
+        created_at=now,
+        updated_at=now,
+    )
+    created = asset_library_repository.create(library)
+    return _to_asset_library_response(created)

apps/api/app/api/routes/assets.py

@@ -7,14 +7,23 @@ from app.dependencies import (
     get_asset_library_repository,
     get_asset_repository,
     get_project_repository,
+    get_tag_repository,
 )
-from app.schemas.asset import AssetResponse, CreateAssetRequest, ListAssetsResponse, UpdateAssetReviewRequest
+from app.schemas.asset import (
+    AssetResponse,
+    BatchDeleteRequest,
+    BatchDeleteResponse,
+    CreateAssetRequest,
+    ListAssetsResponse,
+    UpdateAssetRequest,
+    UpdateAssetReviewRequest,
+)
+from app.schemas.tag import TagAssetsRequest
 from fastapi import APIRouter, Depends, HTTPException, Query
 
 from packages.application import (
     CreateAssetCommand,
     CreateAssetUseCase,
-    ListAssetsUseCase,
 )
 from packages.domain import AssetStatus, ClassificationStatus
 
@@ -59,6 +68,7 @@ def _to_asset_response(item, storage_service=None) -> AssetResponse:
         classification_status=item.classification_status.value,
         quality_score=item.quality_score,
         uploaded_by_user_id=item.uploaded_by_user_id,
+        tag_ids=getattr(item, "tag_ids", []),
     )
 
 
@@ -75,6 +85,13 @@ def _check_project_access(project_id: str, user_id: str, project_repository) ->
 def list_assets(
     library_id: Optional[str] = Query(None),
     project_id: Optional[str] = Query(None),
+    kind: Optional[str] = Query(None, pattern="^(video|voice|image)$"),
+    keyword: Optional[str] = Query(None, description="按名称模糊匹配"),
+    gender: Optional[str] = Query(None, description="按 metadata.gender 筛选"),
+    style: Optional[str] = Query(None, description="按 metadata.style 筛选"),
+    tag_ids: Optional[str] = Query(None, description="按标签 ID 筛选（逗号分隔，取交集）"),
+    skip: int = Query(0, ge=0),
+    limit: int = Query(100, ge=1, le=500),
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     asset_repository: Any = Depends(get_asset_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
@@ -82,32 +99,148 @@ def list_assets(
 ) -> ListAssetsResponse:
     user_id = authenticated_user.user.id
 
-    # 模式1：指定 library_id → 返回该库的素材
+    # kind → file_type 映射（voice 对应 audio）
+    kind_to_file_type = {"video": "video", "voice": "audio", "image": "image"}
+
+    # 解析 tag_ids 参数（逗号分隔）
+    filter_tag_ids: list[str] | None = None
+    if tag_ids:
+        filter_tag_ids = [t.strip() for t in tag_ids.split(",") if t.strip()]
+        if not filter_tag_ids:
+            filter_tag_ids = None
+
+    # 需要内存过滤的标志（keyword/gender/style/tag_ids 无法在 DB 层过滤）
+    needs_memory_filter = bool(keyword or gender or style or filter_tag_ids)
+
+    def _apply_memory_filters(items):
+        """应用 keyword / gender / style / tag_ids 内存过滤。"""
+        result = items
+        if keyword:
+            kw = keyword.lower()
+            result = [i for i in result if kw in (i.name or "").lower()]
+        if gender:
+            result = [i for i in result if (i.metadata or {}).get("gender") == gender]
+        if style:
+            result = [i for i in result if (i.metadata or {}).get("style") == style]
+        if filter_tag_ids:
+            tag_set = set(filter_tag_ids)
+            result = [i for i in result if tag_set.issubset(set(getattr(i, "tag_ids", [])))]
+        return result
+
+    # ── 优化路径：无内存过滤时，使用 DB 级分页 ──
+    if not needs_memory_filter:
+        ft = kind_to_file_type.get(kind) if kind else None
+
+        # 模式1：指定 library_id
+        if library_id:
+            library = asset_library_repository.get(library_id)
+            if library is None:
+                raise HTTPException(status_code=404, detail=f"AssetLibrary {library_id} not found")
+            _check_project_access(library.project_id, user_id, project_repository)
+            if ft:
+                items = asset_repository.find_by_library_and_file_type(library_id, ft, skip=skip, limit=limit)
+                total = asset_repository.count_by_project(library.project_id) if not kind else len(items)
+            else:
+                items = asset_repository.find_by_library(library_id, skip=skip, limit=limit)
+                total = asset_repository.count_by_project(library.project_id)
+            return ListAssetsResponse(
+                items=[_to_asset_response(item) for item in items],
+                total=total,
+                skip=skip,
+                limit=limit,
+            )
+
+        # 模式2：指定 project_id
+        if project_id:
+            _check_project_access(project_id, user_id, project_repository)
+            if ft:
+                # 无直接方法，加载后按 file_type 过滤（仍比全量加载好）
+                all_items = asset_repository.find_by_project(project_id)
+                items = [i for i in all_items if i.mime_type and i.mime_type.startswith(ft)]
+                total = len(items)
+                paged = items[skip : skip + limit]
+            else:
+                items = asset_repository.find_by_project(project_id, skip=skip, limit=limit)
+                total = asset_repository.count_by_project(project_id)
+                paged = items
+            return ListAssetsResponse(
+                items=[_to_asset_response(item) for item in paged],
+                total=total,
+                skip=skip,
+                limit=limit,
+            )
+
+        # 模式3：跨项目（无 library_id/project_id）
+        try:
+            projects = project_repository.find_accessible_projects(user_id)
+        except Exception:
+            logger.exception("查询用户可访问项目失败: user_id=%s", user_id)
+            return ListAssetsResponse(items=[], total=0, skip=skip, limit=limit)
+
+        project_ids = [p.id for p in projects]
+        if not project_ids:
+            return ListAssetsResponse(items=[], total=0, skip=skip, limit=limit)
+
+        total = asset_repository.count_by_project_ids(project_ids)
+        # 跨项目分页：逐项目累积直到凑够一页
+        paged_items: list = []
+        offset = skip
+        remaining = limit
+        for pid in project_ids:
+            proj_total = asset_repository.count_by_project(pid)
+            if offset >= proj_total:
+                offset -= proj_total
+                continue
+            proj_items = asset_repository.find_by_project(pid, skip=offset, limit=remaining)
+            paged_items.extend(proj_items)
+            remaining -= len(proj_items)
+            offset = 0
+            if remaining <= 0:
+                break
+
+        return ListAssetsResponse(
+            items=[_to_asset_response(item) for item in paged_items],
+            total=total,
+            skip=skip,
+            limit=limit,
+        )
+
+    # ── 内存过滤路径：有 keyword/gender/style 时，加载全量后内存过滤 ──
     if library_id:
         library = asset_library_repository.get(library_id)
         if library is None:
             raise HTTPException(status_code=404, detail=f"AssetLibrary {library_id} not found")
         _check_project_access(library.project_id, user_id, project_repository)
-        items = asset_repository.find_by_library(library_id)
-        return ListAssetsResponse(items=[_to_asset_response(item) for item in items])
-
-    # 模式2：指定 project_id → 返回该项目所有素材
-    if project_id:
+        if kind:
+            all_items = asset_repository.find_by_library_and_file_type(library_id, kind_to_file_type[kind])
+        else:
+            all_items = asset_repository.find_by_library(library_id)
+    elif project_id:
         _check_project_access(project_id, user_id, project_repository)
-        items = asset_repository.find_by_project(project_id)
-        return ListAssetsResponse(items=[_to_asset_response(item) for item in items])
+        all_items = asset_repository.find_by_project(project_id)
+    else:
+        try:
+            projects = project_repository.find_accessible_projects(user_id)
+        except Exception:
+            logger.exception("查询用户可访问项目失败: user_id=%s", user_id)
+            return ListAssetsResponse(items=[], total=0, skip=skip, limit=limit)
+        all_items = []
+        for proj in projects:
+            all_items.extend(asset_repository.find_by_project(proj.id))
 
-    # 模式3：都不传 → 返回用户可访问的所有项目的所有素材
-    try:
-        projects = project_repository.find_accessible_projects(user_id)
-    except Exception:
-        logger.exception("查询用户可访问项目失败: user_id=%s", user_id)
-        return ListAssetsResponse(items=[])
-
-    all_items = []
-    for proj in projects:
-        all_items.extend(asset_repository.find_by_project(proj.id))
-    return ListAssetsResponse(items=[_to_asset_response(item) for item in all_items])
+    # 应用 kind 过滤（如果有）+ keyword/gender/style
+    if kind:
+        ft = kind_to_file_type.get(kind)
+        all_items = [i for i in all_items if i.mime_type and i.mime_type.startswith(ft or "")]
+    filtered = _apply_memory_filters(all_items)
+    total = len(filtered)
+    paged = filtered[skip : skip + limit]
+    return ListAssetsResponse(
+        items=[_to_asset_response(item) for item in paged],
+        total=total,
+        skip=skip,
+        limit=limit,
+    )
 
 
 def _apply_asset_review_status(item, review_status: str):
@@ -135,6 +268,130 @@ def update_asset_review_status(
     return _to_asset_response(updated)
 
 
+@router.post("/batch-delete", response_model=BatchDeleteResponse)
+def batch_delete_assets(
+    request: BatchDeleteRequest,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> BatchDeleteResponse:
+    """批量删除素材（配音素材等），需逐项校验项目权限。"""
+    user_id = authenticated_user.user.id
+    deleted_ids: list[str] = []
+    failed_ids: list[str] = []
+
+    for asset_id in request.ids:
+        item = asset_repository.find_by_id(asset_id)
+        if item is None:
+            failed_ids.append(asset_id)
+            continue
+        try:
+            _check_project_access(item.project_id, user_id, project_repository)
+            deleted_ids.append(asset_id)
+        except HTTPException:
+            failed_ids.append(asset_id)
+
+    if deleted_ids:
+        asset_repository.batch_delete(deleted_ids)
+
+    return BatchDeleteResponse(deleted_count=len(deleted_ids), failed_ids=failed_ids)
+
+
+@router.get("/{asset_id}", response_model=AssetResponse)
+def get_asset(
+    asset_id: str,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> AssetResponse:
+    item = asset_repository.find_by_id(asset_id)
+    if item is None:
+        raise HTTPException(status_code=404, detail=f"Asset {asset_id} not found")
+    _check_project_access(item.project_id, authenticated_user.user.id, project_repository)
+    return _to_asset_response(item)
+
+
+@router.put("/{asset_id}", response_model=AssetResponse)
+def update_asset(
+    asset_id: str,
+    request: UpdateAssetRequest,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> AssetResponse:
+    item = asset_repository.find_by_id(asset_id)
+    if item is None:
+        raise HTTPException(status_code=404, detail=f"Asset {asset_id} not found")
+    _check_project_access(item.project_id, authenticated_user.user.id, project_repository)
+
+    # 合并可修改字段
+    if request.name is not None:
+        item.name = request.name
+    if request.metadata is not None:
+        item.metadata = {**item.metadata, **request.metadata}
+    if request.tags is not None:
+        item.metadata = {**item.metadata, "tags": request.tags}
+
+    updated = asset_repository.update(item)
+    return _to_asset_response(updated)
+
+
+@router.delete("/{asset_id}", status_code=204)
+def delete_asset(
+    asset_id: str,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> None:
+    item = asset_repository.find_by_id(asset_id)
+    if item is None:
+        raise HTTPException(status_code=404, detail=f"Asset {asset_id} not found")
+    _check_project_access(item.project_id, authenticated_user.user.id, project_repository)
+    asset_repository.delete(asset_id)
+
+
+@router.post("/{asset_id}/tags", response_model=AssetResponse)
+def tag_asset(
+    asset_id: str,
+    request: TagAssetsRequest,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+    tag_repository: Any = Depends(get_tag_repository),
+) -> AssetResponse:
+    """给素材打标签。"""
+    item = asset_repository.find_by_id(asset_id)
+    if item is None:
+        raise HTTPException(status_code=404, detail=f"Asset {asset_id} not found")
+    _check_project_access(item.project_id, authenticated_user.user.id, project_repository)
+    for tag_id in request.tag_ids:
+        tag = tag_repository.get(tag_id)
+        if tag is None:
+            raise HTTPException(status_code=404, detail=f"Tag {tag_id} not found")
+        if tag.user_id != authenticated_user.user.id:
+            raise HTTPException(status_code=403, detail=f"无权使用标签 {tag_id}")
+        item.add_tag(tag_id)
+    updated = asset_repository.update(item)
+    return _to_asset_response(updated)
+
+
+@router.delete("/{asset_id}/tags/{tag_id}", status_code=204)
+def untag_asset(
+    asset_id: str,
+    tag_id: str,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    asset_repository: Any = Depends(get_asset_repository),
+    project_repository: Any = Depends(get_project_repository),
+) -> None:
+    """取消素材的标签。"""
+    item = asset_repository.find_by_id(asset_id)
+    if item is None:
+        raise HTTPException(status_code=404, detail=f"Asset {asset_id} not found")
+    _check_project_access(item.project_id, authenticated_user.user.id, project_repository)
+    item.remove_tag(tag_id)
+    asset_repository.update(item)
+
+
 @router.post("", response_model=AssetResponse)
 def create_asset(
     request: CreateAssetRequest,

apps/api/app/api/routes/auth.py

@@ -6,13 +6,14 @@ app.dependencies and authentication behavior lives in application use cases.
 """
 
 import logging
+import os
 from typing import Optional
 
 import jwt
 from app.auth import AuthenticatedUser, blacklist_token, get_current_user
 from app.config import settings
 from app.dependencies import get_auth_email_service, get_auth_session_store, get_user_repository
-from fastapi import APIRouter, Depends, HTTPException, status
+from fastapi import APIRouter, Depends, Header, HTTPException, status
 from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 from pydantic import BaseModel, EmailStr
 
@@ -284,3 +285,99 @@ def _translate_auth_error(error: str | None) -> str:
         "Display name is required": "显示名称不能为空",
     }
     return translations.get(error or "", error or "注册失败")
+
+
+class WechatSyncRequest(BaseModel):
+    openid: str
+    unionid: Optional[str] = None
+    nickname: Optional[str] = None
+    avatar_url: Optional[str] = None
+    source: str = "miniapp"
+
+
+class WechatSyncResponse(BaseModel):
+    access_token: str
+    token: str
+    refresh_token: str
+    user_id: str
+    user: dict
+    user_info: dict
+    is_new_user: bool
+    expires_in: int
+
+
+def _get_internal_api_keys() -> list[str]:
+    """获取内部 API Key 列表
+
+    优先级:
+    1. INTERNAL_API_KEYS 环境变量
+    2. /app/generated/internal_api_keys.txt 文件 (volume 持久化)
+    """
+    env_keys = os.environ.get("INTERNAL_API_KEYS", "")
+    if env_keys:
+        return [k.strip() for k in env_keys.split(",") if k.strip()]
+
+    # 从持久化文件读取
+    try:
+        with open("/app/generated/internal_api_keys.txt", "r") as f:
+            content = f.read().strip()
+            if content:
+                return [k.strip() for k in content.split(",") if k.strip()]
+    except Exception:
+        logger.debug("Failed to read internal API keys from file", exc_info=True)
+
+    return []
+
+
+def _verify_internal_api_key(x_api_key: str | None = Header(None)) -> bool:
+    """验证内部 API Key
+
+    - 已配置时：必须匹配 INTERNAL_API_KEYS 中的 key
+    - 未配置且非生产环境：放行（方便开发）
+    - 未配置且生产环境：拒绝
+    """
+    env = os.environ.get("APP_ENV", os.environ.get("ENV", "development")).lower()
+    key_list = _get_internal_api_keys()
+
+    if not key_list:
+        if env in ("production", "prod"):
+            raise HTTPException(status_code=401, detail="内部接口未配置 API Key")
+        return True
+
+    if x_api_key and x_api_key.strip() in key_list:
+        return True
+
+    raise HTTPException(status_code=401, detail="无效的 API Key")
+
+
+@router.post("/wechat-sync", response_model=WechatSyncResponse, include_in_schema=False)
+async def wechat_sync(
+    request: WechatSyncRequest,
+    user_repository: UserRepository = Depends(get_user_repository),
+    _: bool = Depends(_verify_internal_api_key),
+):
+    """
+    微信同步登录/注册（系统级内部接口）
+
+    由 BFF 层通过 API Key 调用，不直接面向终端用户。
+    根据 openid 查找或创建用户，返回 SaaS token。
+    """
+    from packages.application.auth.wechat_sync_use_case import WechatSyncRequest as UseCaseRequest
+    from packages.application.auth.wechat_sync_use_case import (
+        WechatSyncUseCase,
+    )
+
+    use_case = WechatSyncUseCase(user_repository=user_repository)
+    use_case_request = UseCaseRequest(
+        openid=request.openid,
+        unionid=request.unionid,
+        nickname=request.nickname,
+        avatar_url=request.avatar_url,
+        source=request.source,
+    )
+
+    response, error = use_case.execute(use_case_request)
+    if error:
+        raise HTTPException(status_code=400, detail=error)
+
+    return WechatSyncResponse(**response.to_dict())

apps/api/app/api/routes/chunked_upload.py

@@ -19,6 +19,7 @@ from app.core.celery_app import celery_app
 from app.core.storage import OSSStorageService, get_storage_service
 from app.dependencies import (
     get_asset_library_repository,
+    get_asset_repository,
     get_ingest_job_repository,
     get_project_repository,
 )
@@ -42,20 +43,33 @@ DEFAULT_CHUNK_SIZE = 5 * 1024 * 1024  # 5MB
 MAX_FILE_SIZE = 2 * 1024 * 1024 * 1024  # 2GB
 CHUNK_EXPIRY_HOURS = 24
 
-# Allowed file types (consistent with existing upload.py)
+# Allowed file types — must stay in sync with upload.py ALLOWED_MIME_TYPES
 ALLOWED_MIME_TYPES = {
+    # Images
     "image/jpeg",
     "image/png",
     "image/gif",
     "image/webp",
+    "image/bmp",
+    "image/tiff",
+    "image/svg+xml",
+    # Video
     "video/mp4",
     "video/quicktime",
+    "video/mpeg",
     "video/x-msvideo",
     "video/webm",
+    "video/x-matroska",
+    "video/3gpp",
+    # Audio
     "audio/mpeg",
     "audio/wav",
     "audio/ogg",
     "audio/mp3",
+    "audio/flac",
+    "audio/aac",
+    "audio/x-m4a",
+    "audio/webm",
 }
 
 # Chunk storage root directory
@@ -360,6 +374,7 @@ async def complete_chunked_upload(
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     project_repository: Any = Depends(get_project_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
+    asset_repository: Any = Depends(get_asset_repository),
     ingest_job_repository: Any = Depends(get_ingest_job_repository),
     storage_service: OSSStorageService = Depends(get_storage_service),
 ) -> ChunkedUploadCompleteResponse:
@@ -423,6 +438,29 @@ async def complete_chunked_upload(
             content_type=meta["content_type"],
         )
 
+        # ── 素材去重检测：同素材库 + 同 file_hash 视为重复 ──
+        if request.file_hash:
+            existing = asset_repository.find_by_library_and_file_hash(
+                library_id=request.library_id,
+                file_hash=request.file_hash,
+            )
+            if existing is not None:
+                logger.info(
+                    "素材去重命中(chunked): library=%s hash=%s existing_asset=%s",
+                    request.library_id,
+                    request.file_hash,
+                    existing.id,
+                )
+                meta["status"] = "completed"
+                _save_upload_meta(upload_id, meta)
+                return ChunkedUploadCompleteResponse(
+                    storage_key=storage_key,
+                    ingest_job_id="",
+                    url=file_url,
+                    duplicated=True,
+                    asset_id=existing.id,
+                )
+
         # Create ingest job
         use_case = SubmitIngestJobUseCase(ingest_job_repository)
         job = use_case.execute(
@@ -430,6 +468,7 @@ async def complete_chunked_upload(
                 project_id=meta["project_id"],
                 library_id=meta["library_id"],
                 storage_key=storage_key,
+                file_hash=request.file_hash,
             )
         )
         celery_app.send_task("worker.ingest_asset", args=[job.id])

apps/api/app/api/routes/generation_tasks.py

@@ -1,3 +1,5 @@
+import random
+import uuid
 from typing import Any
 
 from app.auth import AuthenticatedUser, get_current_user
@@ -14,6 +16,7 @@ from app.schemas.generated_video import (
     ListGeneratedVideosResponse,
 )
 from app.schemas.generation_task import (
+    BatchGenerationTaskResponse,
     CreateGenerationTaskRequest,
     GenerationTaskResponse,
     ListGenerationTasksResponse,
@@ -51,6 +54,8 @@ def _to_generation_task_response(task) -> GenerationTaskResponse:
         title_ids=task.title_ids,
         voice_ids=task.voice_ids,
         source_edit_plan_id=task.source_edit_plan_id or "",
+        asset_select_mode=getattr(task, "asset_select_mode", ""),
+        batch_id=getattr(task, "batch_id", ""),
         status=task.status,
         progress=task.progress,
         result_count=task.result_count,
@@ -85,6 +90,49 @@ def _ensure_library_has_ready_video_assets(assets) -> None:
         )
 
 
+def _select_assets_from_library(
+    assets: list,
+    mode: str,
+    count: int,
+) -> list[str]:
+    """根据选取模式从素材库中选取 ready 状态的视频素材 ID。
+
+    Args:
+        assets: 素材库中所有素材（Asset 实体列表）
+        mode: 选取模式 — all=全部, random=随机, smart=按质量评分
+        count: 选取数量，0 表示全部（仅 random/smart 模式有效）
+
+    Returns:
+        选中的素材 ID 列表
+    """
+    ready_video_assets = [a for a in assets if a.status.value == "ready" and a.mime_type.startswith("video")]
+
+    if not ready_video_assets:
+        return []
+
+    if mode == "random":
+        selected = (
+            ready_video_assets if count <= 0 else random.sample(ready_video_assets, min(count, len(ready_video_assets)))
+        )
+        return [a.id for a in selected]
+
+    if mode == "smart":
+        # 按质量分降序排列（质量分高的优先），质量分相同时按时长降序
+        sorted_assets = sorted(
+            ready_video_assets,
+            key=lambda a: (
+                a.quality_score if a.quality_score is not None else 0.0,
+                a.duration if a.duration is not None else 0.0,
+            ),
+            reverse=True,
+        )
+        selected = sorted_assets if count <= 0 else sorted_assets[:count]
+        return [a.id for a in selected]
+
+    # 默认 all 模式：返回全部 ready 视频素材
+    return [a.id for a in ready_video_assets]
+
+
 def _resolve_project_and_library(
     request: CreateGenerationTaskRequest,
     project_repository: Any,
@@ -122,7 +170,7 @@ def _resolve_project_and_library(
     return project_id, asset_library_id
 
 
-@router.post("/tasks", response_model=GenerationTaskResponse)
+@router.post("/tasks", response_model=BatchGenerationTaskResponse)
 def create_generation_task(
     request: CreateGenerationTaskRequest,
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
@@ -130,12 +178,13 @@ def create_generation_task(
     project_repository: Any = Depends(get_project_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
     asset_repository: Any = Depends(get_asset_repository),
-) -> GenerationTaskResponse:
+) -> BatchGenerationTaskResponse:
     project_id, asset_library_id = _resolve_project_and_library(
         request, project_repository, asset_library_repository, asset_repository, authenticated_user
     )
 
     # asset_library 存在性校验（仅在提供了 asset_library_id 时）
+    resolved_asset_ids: list[str] = list(request.asset_ids)
     if asset_library_id:
         library = asset_library_repository.get(asset_library_id)
         if library is None or (project_id and library.project_id != project_id):
@@ -144,23 +193,42 @@ def create_generation_task(
         assets = asset_repository.find_by_library(asset_library_id)
         _ensure_library_has_ready_video_assets(assets)
 
+        # 素材库自动匹配：当未显式指定 asset_ids 时，按模式自动选取
+        if not resolved_asset_ids:
+            resolved_asset_ids = _select_assets_from_library(
+                assets,
+                mode=request.asset_select_mode,
+                count=request.asset_select_count,
+            )
+
     use_case = CreateGenerationTaskUseCase(generation_task_repository)
-    task = use_case.execute(
-        CreateGenerationTaskCommand(
-            project_id=project_id,
-            asset_library_id=asset_library_id,
-            strategy_id=request.strategy_id,
-            voice_library_id=request.voice_library_id,
-            template_id=request.template_id,
-            asset_ids=request.asset_ids,
-            title_ids=request.title_ids,
-            voice_ids=request.voice_ids,
-            created_by_user_id=authenticated_user.user.id,
-            source_edit_plan_id=request.source_edit_plan_id,
+    count = request.count
+    created_tasks = []
+    # 同批次任务共享 batch_id，用于视频查重时批次内比对
+    batch_id = uuid.uuid4().hex if count > 1 else ""
+
+    for _ in range(count):
+        task = use_case.execute(
+            CreateGenerationTaskCommand(
+                project_id=project_id,
+                asset_library_id=asset_library_id,
+                strategy_id=request.strategy_id,
+                voice_library_id=request.voice_library_id,
+                template_id=request.template_id,
+                asset_ids=resolved_asset_ids,
+                title_ids=request.title_ids,
+                voice_ids=request.voice_ids,
+                created_by_user_id=authenticated_user.user.id,
+                source_edit_plan_id=request.source_edit_plan_id,
+                asset_select_mode=request.asset_select_mode,
+                batch_id=batch_id,
+            )
         )
-    )
-    celery_app.send_task("worker.generate_video", args=[task.id])
-    return _to_generation_task_response(task)
+        celery_app.send_task("worker.generate_video", args=[task.id])
+        created_tasks.append(task)
+
+    items = [_to_generation_task_response(t) for t in created_tasks]
+    return BatchGenerationTaskResponse(items=items, total=len(items))
 
 
 @router.get("/tasks", response_model=ListGenerationTasksResponse)
@@ -237,6 +305,7 @@ def retry_generation_task(
             voice_ids=task.voice_ids,
             created_by_user_id=authenticated_user.user.id,
             source_edit_plan_id=task.source_edit_plan_id or "",
+            asset_select_mode=getattr(task, "asset_select_mode", ""),
         )
     )
     celery_app.send_task("worker.generate_video", args=[retried.id])

apps/api/app/api/routes/tags.py

@@ -0,0 +1,67 @@
+"""标签 CRUD 路由。"""
+
+import logging
+from typing import Any
+
+from app.auth import AuthenticatedUser, get_current_user
+from app.dependencies import get_tag_repository
+from app.schemas.tag import (
+    CreateTagRequest,
+    ListTagsResponse,
+    TagResponse,
+)
+from fastapi import APIRouter, Depends, HTTPException
+
+from packages.domain import Tag
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter()
+
+
+@router.get("", response_model=ListTagsResponse)
+def list_tags(
+    skip: int = 0,
+    limit: int = 100,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    tag_repository: Any = Depends(get_tag_repository),
+) -> ListTagsResponse:
+    """列出当前用户的标签。"""
+    user_id = authenticated_user.user.id
+    items = tag_repository.list_by_user(user_id, skip=skip, limit=limit)
+    total = tag_repository.count_by_user(user_id)
+    return ListTagsResponse(
+        items=[TagResponse(id=t.id, name=t.name, created_at=t.created_at) for t in items],
+        total=total,
+    )
+
+
+@router.post("", response_model=TagResponse, status_code=201)
+def create_tag(
+    request: CreateTagRequest,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    tag_repository: Any = Depends(get_tag_repository),
+) -> TagResponse:
+    """创建标签（同用户同名去重，返回 409）。"""
+    user_id = authenticated_user.user.id
+    existing = tag_repository.find_by_name(user_id, request.name)
+    if existing:
+        raise HTTPException(status_code=409, detail="标签名称已存在")
+    tag = Tag.create(user_id=user_id, name=request.name)
+    created = tag_repository.create(tag)
+    return TagResponse(id=created.id, name=created.name, created_at=created.created_at)
+
+
+@router.delete("/{tag_id}", status_code=204)
+def delete_tag(
+    tag_id: str,
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    tag_repository: Any = Depends(get_tag_repository),
+) -> None:
+    """删除标签（同时清理素材关联）。"""
+    tag = tag_repository.get(tag_id)
+    if tag is None:
+        raise HTTPException(status_code=404, detail="标签不存在")
+    if tag.user_id != authenticated_user.user.id:
+        raise HTTPException(status_code=403, detail="无权删除该标签")
+    tag_repository.delete(tag_id)

apps/api/app/api/routes/tts.py

@@ -6,21 +6,31 @@ import logging
 from typing import Optional
 
 from app.auth import AuthenticatedUser, get_current_user
-from app.dependencies import get_cosyvoice_service, get_db_session
+from app.dependencies import (
+    get_cosyvoice_service,
+    get_db_session,
+    get_user_repository,
+    get_voice_clone_profile_repository,
+    get_voice_library_repository,
+)
 from app.schemas.tts import (
     ListTTSJobResponse,
+    SaveToLibraryRequest,
+    SaveToLibraryResponse,
     TTSJobResponse,
     TTSStatusResponse,
     TTSSynthesizeRequest,
     TTSSynthesizeResponse,
 )
-from fastapi import APIRouter, Depends, HTTPException, Query, Response, status
+from fastapi import APIRouter, Depends, HTTPException, Query, Response, WebSocket, WebSocketDisconnect, status
 from sqlalchemy.orm import Session
 
 from packages.adapters.sqlalchemy_impl.tts_job_repository import (
     SQLAlchemyTTSJobRepository,
 )
+from packages.adapters.sqlalchemy_impl.voice_library_repository import SQLAlchemyVoiceLibraryRepository
 from packages.application.cosyvoice_service import CosyVoiceService
+from packages.application.tts_job.streaming_service import TTSStreamingService
 from packages.application.tts_job.use_cases import (
     CreateTTSJobUseCase,
     DeleteTTSJobUseCase,
@@ -30,6 +40,12 @@ from packages.application.tts_job.use_cases import (
     TTSJobNotFoundError,
 )
 from packages.application.tts_job.workflow import TTSWorkflowService
+from packages.application.voice_library.commands import CreateVoiceLibraryCommand
+from packages.application.voice_library.use_cases import (
+    CreateVoiceLibraryUseCase,
+    QuotaExceededError,
+)
+from packages.ports.user_repository import UserRepository
 
 logger = logging.getLogger(__name__)
 
@@ -73,6 +89,7 @@ def synthesize(
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     repository: SQLAlchemyTTSJobRepository = Depends(_get_repository),
     cosyvoice_service: CosyVoiceService = Depends(get_cosyvoice_service),
+    voice_clone_repo=Depends(get_voice_clone_profile_repository),
 ) -> TTSSynthesizeResponse:
     """发起 TTS 合成任务。
 
@@ -80,6 +97,21 @@ def synthesize(
     与音色克隆接口保持一致：CosyVoice 失败时不抛 500，而是返回 201 + failed 状态任务记录。
     """
     user_id = authenticated_user.user.id
+
+    # 校验 voice_clone_profile_id 归属（防止越权使用他人克隆音色）
+    if request.voice_clone_profile_id:
+        profile = voice_clone_repo.get(request.voice_clone_profile_id)
+        if profile is None:
+            raise HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND,
+                detail="Voice clone profile not found",
+            )
+        if profile.user_id != user_id:
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="Access denied to voice clone profile",
+            )
+
     use_case = CreateTTSJobUseCase(repository)
     job = use_case.execute(
         user_id=user_id,
@@ -110,18 +142,25 @@ def synthesize(
 
     # 若任务处于 processing 状态（异步模式），触发 Celery 后台轮询
     if job.status.value == "processing":
-        task_id = (job.metadata or {}).get("cosyvoice_task_id", "")
-        if task_id:
-            try:
+        # 分段合成任务 vs 普通单段任务
+        segment_task_ids = (job.metadata or {}).get("segment_task_ids", [])
+        is_segment = len(segment_task_ids) > 0
+
+        try:
+            if is_segment:
+                from worker_app.tasks import process_tts_segment_synthesis
+
+                process_tts_segment_synthesis.delay(job.id)
+            else:
                 from worker_app.tasks import process_tts_synthesis
 
                 process_tts_synthesis.delay(job.id)
-            except Exception as e:
-                # Celery 调度失败，标记 job 为 failed
-                try:
-                    workflow.process_synthesis_failure(job.id, f"Celery 任务调度失败: {e}")
-                except Exception as inner_e:
-                    logger.error(f"Celery 调度后标记失败时出错: job_id={job.id}, error={inner_e}")
+        except Exception as e:
+            # Celery 调度失败，标记 job 为 failed
+            try:
+                workflow.process_synthesis_failure(job.id, f"Celery 任务调度失败: {e}")
+            except Exception as inner_e:
+                logger.error(f"Celery 调度后标记失败时出错: job_id={job.id}, error={inner_e}")
 
     return TTSSynthesizeResponse(
         job_id=job.id,
@@ -205,3 +244,127 @@ def delete_tts_job(
     if not deleted:
         raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="TTS job not found")
     return Response(status_code=204)
+
+
+@router.post(
+    "/jobs/{job_id}/save-to-library",
+    response_model=SaveToLibraryResponse,
+    status_code=status.HTTP_201_CREATED,
+)
+def save_tts_job_to_library(
+    job_id: str,
+    request: SaveToLibraryRequest = SaveToLibraryRequest(),
+    authenticated_user: AuthenticatedUser = Depends(get_current_user),
+    tts_repository: SQLAlchemyTTSJobRepository = Depends(_get_repository),
+    voice_library_repository: SQLAlchemyVoiceLibraryRepository = Depends(get_voice_library_repository),
+    user_repository: UserRepository = Depends(get_user_repository),
+) -> SaveToLibraryResponse:
+    """将已完成的 TTS 合成结果保存到配音库。
+
+    自动携带音色名、时长、语速等元信息。
+    """
+    user_id = authenticated_user.user.id
+
+    # 获取 TTS job
+    get_use_case = GetTTSJobUseCase(tts_repository)
+    try:
+        job = get_use_case.execute(job_id, user_id)
+    except TTSJobNotFoundError:
+        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="TTS job not found")
+
+    # 校验已完成
+    if not job.is_completed:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail="TTS job is not completed yet",
+        )
+
+    # 构建配音素材名称
+    name = request.name or f"TTS-{job.id[:8]}"
+
+    # 构建元信息
+    metadata_ = {
+        "source": "tts_job",
+        "tts_job_id": job.id,
+        "format": job.format,
+        "sample_rate": job.sample_rate,
+    }
+    if job.metadata:
+        # 保留原始 job 的有用元信息
+        for key in ("speed", "language"):
+            if key in job.metadata:
+                metadata_[key] = job.metadata[key]
+
+    # 获取用户套餐（用于配额检查）
+    user = user_repository.find_by_id(user_id)
+    plan_name = getattr(user, "subscription_plan", "free") if user else "free"
+
+    # 构建命令并执行
+    command = CreateVoiceLibraryCommand(
+        user_id=user_id,
+        name=name,
+        text=job.input_text,
+        voice_provider="cosyvoice",
+        voice_id=job.voice_id,
+        voice_name=job.voice_model or "",
+        audio_url=job.output_audio_url,
+        duration=job.duration,
+        file_size=job.file_size,
+        status="completed",
+        project_id=job.project_id or "",
+        tags=[],
+        metadata_=metadata_,
+    )
+
+    use_case = CreateVoiceLibraryUseCase(voice_library_repository)
+    try:
+        item = use_case.execute(command, plan_name=plan_name or "free")
+    except QuotaExceededError as exc:
+        raise HTTPException(
+            status_code=status.HTTP_429_TOO_MANY_REQUESTS,
+            detail=f"配音库配额已满（{exc.used}/{exc.limit}），请升级套餐",
+        )
+
+    return SaveToLibraryResponse(
+        id=item.id,
+        name=item.name,
+        audio_url=item.audio_url,
+        duration=item.duration,
+        voice_id=item.voice_id,
+        voice_name=item.voice_name,
+        status=item.status,
+    )
+
+
+@router.websocket("/ws/tts/stream")
+async def tts_websocket_stream(
+    websocket: WebSocket,
+    cosyvoice_service: CosyVoiceService = Depends(get_cosyvoice_service),
+) -> None:
+    """WebSocket 流式 TTS 合成。
+
+    协议：
+    1. 客户端发送 JSON 文本帧: {"text": "...", "voice_id": "...", ...}
+    2. 服务端发送 JSON 状态帧 + 二进制音频帧
+    3. 完成时发送 JSON 结束帧
+    """
+    await websocket.accept()
+    try:
+        message = await websocket.receive_json()
+        params = {
+            "text": message.get("text", ""),
+            "voice_id": message.get("voice_id", ""),
+            "sample_rate": message.get("sample_rate", 0),
+            "format": message.get("format", "mp3"),
+            "speed": message.get("speed", 1.0),
+        }
+        streaming_service = TTSStreamingService(cosyvoice_service)
+        await streaming_service.synthesize_and_stream(websocket, params)
+    except WebSocketDisconnect:
+        logger.info("WebSocket 客户端断开连接")
+    except Exception as e:
+        logger.error(f"WebSocket 流式合成异常: {e}", exc_info=True)
+        try:
+            await websocket.send_json({"type": "error", "message": f"服务异常: {e}"})
+        except Exception:
+            pass

apps/api/app/api/routes/upload.py

@@ -8,6 +8,7 @@ from app.core.celery_app import celery_app
 from app.core.storage import OSSStorageService, get_storage_service
 from app.dependencies import (
     get_asset_library_repository,
+    get_asset_repository,
     get_ingest_job_repository,
     get_project_repository,
 )
@@ -99,6 +100,7 @@ def _submit_ingest_job(
     library_id: str,
     storage_key: str,
     ingest_job_repository: Any,
+    file_hash: str = "",
 ) -> Any:
     use_case = SubmitIngestJobUseCase(ingest_job_repository)
     job = use_case.execute(
@@ -106,6 +108,7 @@ def _submit_ingest_job(
             project_id=project_id,
             library_id=library_id,
             storage_key=storage_key,
+            file_hash=file_hash,
         )
     )
     celery_app.send_task("worker.ingest_asset", args=[job.id])
@@ -176,6 +179,7 @@ async def complete_direct_upload(
     ingest_job_repository: Any = Depends(get_ingest_job_repository),
     project_repository: Any = Depends(get_project_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
+    asset_repository: Any = Depends(get_asset_repository),
     storage_service: OSSStorageService = Depends(get_storage_service),
 ) -> DirectUploadCompleteResponse:
     """确认浏览器直传完成并创建导入任务。"""
@@ -199,11 +203,32 @@ async def complete_direct_upload(
     if not file_exists:
         raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="Uploaded file not found")
 
+    # ── 素材去重检测：同素材库 + 同 file_hash 视为重复 ──
+    if request.file_hash:
+        existing = asset_repository.find_by_library_and_file_hash(
+            library_id=request.library_id,
+            file_hash=request.file_hash,
+        )
+        if existing is not None:
+            logger.info(
+                "素材去重命中: library=%s hash=%s existing_asset=%s",
+                request.library_id,
+                request.file_hash,
+                existing.id,
+            )
+            return DirectUploadCompleteResponse(
+                storage_key=normalized_key,
+                ingest_job_id="",
+                duplicated=True,
+                asset_id=existing.id,
+            )
+
     job = _submit_ingest_job(
         project_id=request.project_id,
         library_id=request.library_id,
         storage_key=normalized_key,
         ingest_job_repository=ingest_job_repository,
+        file_hash=request.file_hash,
     )
     return DirectUploadCompleteResponse(storage_key=normalized_key, ingest_job_id=job.id)
 
@@ -218,15 +243,38 @@ async def upload_asset(
     project_id: str = Form(..., min_length=1, description="项目 ID"),
     library_id: str = Form(..., min_length=1, description="素材库 ID"),
     file: UploadFile = File(..., description="要上传的文件（视频、音频、图片等）"),
+    file_hash: str = Form(default="", description="文件 MD5 哈希，用于去重检测"),
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     ingest_job_repository: Any = Depends(get_ingest_job_repository),
     project_repository: Any = Depends(get_project_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
+    asset_repository: Any = Depends(get_asset_repository),
     storage_service: OSSStorageService = Depends(get_storage_service),
 ) -> UploadAssetResponse:
     """上传素材文件并触发导入流水线。"""
     _require_project_and_library(project_id, library_id, project_repository, asset_library_repository)
 
+    # ── 素材去重检测：上传前检查同素材库 + 同 file_hash ──
+    if file_hash:
+        existing = asset_repository.find_by_library_and_file_hash(
+            library_id=library_id,
+            file_hash=file_hash,
+        )
+        if existing is not None:
+            logger.info(
+                "素材去重命中(multipart): library=%s hash=%s existing_asset=%s",
+                library_id,
+                file_hash,
+                existing.id,
+            )
+            return UploadAssetResponse(
+                storage_key=existing.storage_key,
+                ingest_job_id="",
+                url="",
+                duplicated=True,
+                asset_id=existing.id,
+            )
+
     # P2-5: 服务端验证 MIME 类型
     validated_content_type = _validate_mime_type(file.content_type)
 
@@ -255,6 +303,7 @@ async def upload_asset(
         library_id=library_id,
         storage_key=storage_key,
         ingest_job_repository=ingest_job_repository,
+        file_hash=file_hash,
     )
 
     return UploadAssetResponse(

apps/api/app/dependencies.py

@@ -39,6 +39,7 @@ from packages.adapters.sqlalchemy_impl.project_repository import (
     SQLAlchemyProjectRepository,
 )
 from packages.adapters.sqlalchemy_impl.session import build_session_factory
+from packages.adapters.sqlalchemy_impl.tag_repository import SQLAlchemyTagRepository
 from packages.adapters.sqlalchemy_impl.title_library_repository import (
     SQLAlchemyTitleLibraryRepository,
 )
@@ -58,6 +59,7 @@ from packages.ports.generation_task_repository import GenerationTaskRepository
 from packages.ports.ingest_job_repository import IngestJobRepository
 from packages.ports.job_repository import JobRepository
 from packages.ports.project_repository import ProjectRepository
+from packages.ports.tag_repository import TagRepository
 from packages.ports.title_library_repository import TitleLibraryRepository
 from packages.ports.user_repository import UserRepository
 from packages.ports.voice_clone_profile_repository import VoiceCloneProfileRepository
@@ -138,6 +140,13 @@ def get_project_repository(
     return SQLAlchemyProjectRepository(session)
 
 
+def get_tag_repository(
+    session: Session = Depends(get_db_session),
+) -> TagRepository:
+    """Provide the SQLAlchemy tag repository implementation."""
+    return SQLAlchemyTagRepository(session)
+
+
 def get_user_repository(
     session: Session = Depends(get_db_session),
 ) -> UserRepository:

apps/api/app/schemas/asset.py

@@ -25,6 +25,12 @@ class UpdateAssetReviewRequest(BaseModel):
     review_status: str = Field(..., pattern="^(pending_review|approved|rejected)$")
 
 
+class UpdateAssetRequest(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=100)
+    metadata: dict[str, object] | None = None
+    tags: list[str] | None = None
+
+
 class AssetResponse(BaseModel):
     id: str
     project_id: str
@@ -45,7 +51,24 @@ class AssetResponse(BaseModel):
     classification_status: str
     quality_score: float | None = None
     uploaded_by_user_id: str
+    tag_ids: list[str] = Field(default_factory=list)
+
+
+class BatchDeleteRequest(BaseModel):
+    """批量删除请求。"""
+
+    ids: list[str] = Field(..., min_length=1, max_length=100, description="要删除的素材 ID 列表")
+
+
+class BatchDeleteResponse(BaseModel):
+    """批量删除响应。"""
+
+    deleted_count: int = Field(..., ge=0, description="实际删除数量")
+    failed_ids: list[str] = Field(default_factory=list, description="删除失败的 ID 列表")
 
 
 class ListAssetsResponse(BaseModel):
     items: list[AssetResponse]
+    total: int = Field(default=0, ge=0)
+    skip: int = Field(default=0, ge=0)
+    limit: int = Field(default=100, ge=1)

apps/api/app/schemas/asset_library.py

@@ -18,3 +18,8 @@ class AssetLibraryResponse(BaseModel):
 
 class ListAssetLibrariesResponse(BaseModel):
     items: list[AssetLibraryResponse]
+
+
+class EnsureDefaultLibraryRequest(BaseModel):
+    project_id: str = Field(..., min_length=1)
+    kind: str = Field(..., pattern="^(video|voice|image)$")

apps/api/app/schemas/chunked_upload.py

@@ -36,9 +36,12 @@ class ChunkedUploadStatusResponse(BaseModel):
 class ChunkedUploadCompleteRequest(BaseModel):
     project_id: str = Field(..., min_length=1, description="Project ID")
     library_id: str = Field(..., min_length=1, description="Asset library ID")
+    file_hash: str = Field(default="", max_length=64, description="文件 MD5 哈希，用于去重检测")
 
 
 class ChunkedUploadCompleteResponse(BaseModel):
     storage_key: str = Field(..., description="Storage key")
     ingest_job_id: str = Field(..., description="Ingest job ID")
     url: str = Field(..., description="File URL")
+    duplicated: bool = Field(default=False, description="是否为重复素材（命中去重）")
+    asset_id: str = Field(default="", description="重复素材的 asset_id（duplicated=true 时返回）")

apps/api/app/schemas/generation_task.py

@@ -21,6 +21,16 @@ class CreateGenerationTaskRequest(BaseModel):
     voice_ids: list[str] = Field(default_factory=list)
     # ── 来源剪辑计划 ──
     source_edit_plan_id: str = ""
+    # ── 批量生成 ──
+    count: int = Field(default=1, ge=1, le=50, description="批量生成数量，默认1，最大50")
+    # ── 素材库自动匹配 ──
+    asset_select_mode: str = Field(
+        default="all",
+        description="素材选取模式：all=全部ready视频, random=随机选取, smart=智能匹配（按质量/时长评分）",
+    )
+    asset_select_count: int = Field(
+        default=0, ge=0, le=100, description="选取数量，0表示全部（仅 random/smart 模式有效）"
+    )
 
     @model_validator(mode="after")
     def _check_at_least_one_mode(self) -> "CreateGenerationTaskRequest":
@@ -46,12 +56,21 @@ class GenerationTaskResponse(BaseModel):
     title_ids: list[str] = Field(default_factory=list)
     voice_ids: list[str] = Field(default_factory=list)
     source_edit_plan_id: str = ""
+    asset_select_mode: str = ""
+    batch_id: str = ""
     status: str
     progress: float
     result_count: int
     error_message: str
 
 
+class BatchGenerationTaskResponse(BaseModel):
+    """批量生成任务响应。"""
+
+    items: list[GenerationTaskResponse]
+    total: int
+
+
 class ListGenerationTasksResponse(BaseModel):
     """用户级生成任务列表响应（跨 project）。"""
 

apps/api/app/schemas/tag.py

@@ -0,0 +1,24 @@
+"""标签相关 Schema。"""
+
+from datetime import datetime
+
+from pydantic import BaseModel, Field
+
+
+class CreateTagRequest(BaseModel):
+    name: str = Field(..., min_length=1, max_length=100)
+
+
+class TagResponse(BaseModel):
+    id: str
+    name: str
+    created_at: datetime
+
+
+class ListTagsResponse(BaseModel):
+    items: list[TagResponse]
+    total: int = Field(default=0, ge=0)
+
+
+class TagAssetsRequest(BaseModel):
+    tag_ids: list[str] = Field(..., min_length=1, max_length=50)

apps/api/app/schemas/tts.py

@@ -83,3 +83,21 @@ class ListTTSJobResponse(BaseModel):
     total: int
     page: int
     page_size: int
+
+
+class SaveToLibraryRequest(BaseModel):
+    """保存到配音库请求。"""
+
+    name: Optional[str] = Field(None, description="配音素材名称，留空则自动生成")
+
+
+class SaveToLibraryResponse(BaseModel):
+    """保存到配音库响应。"""
+
+    id: str
+    name: str
+    audio_url: str
+    duration: float
+    voice_id: str
+    voice_name: str
+    status: str

apps/api/app/schemas/upload.py

@@ -6,12 +6,7 @@ class UploadAssetRequest(BaseModel):
 
     project_id: str = Field(..., min_length=1, description="项目 ID")
     library_id: str = Field(..., min_length=1, description="素材库 ID")
-
-
-class UploadAssetResponse(BaseModel):
-    storage_key: str
-    ingest_job_id: str
-    url: str = Field(..., description="Public URL of uploaded file")
+    file_hash: str = Field(default="", max_length=64, description="文件 MD5 哈希，用于去重检测")
 
 
 class DirectUploadPrepareRequest(BaseModel):
@@ -20,6 +15,7 @@ class DirectUploadPrepareRequest(BaseModel):
     filename: str = Field(..., min_length=1, max_length=255)
     content_type: str = Field(default="application/octet-stream", min_length=1, max_length=100)
     file_size: int = Field(..., gt=0)
+    file_hash: str = Field(default="", max_length=64, description="文件 MD5 哈希，用于去重检测")
 
 
 class DirectUploadPrepareResponse(BaseModel):
@@ -35,8 +31,19 @@ class DirectUploadCompleteRequest(BaseModel):
     project_id: str = Field(..., min_length=1)
     library_id: str = Field(..., min_length=1)
     storage_key: str = Field(..., min_length=1, max_length=255)
+    file_hash: str = Field(default="", max_length=64, description="文件 MD5 哈希，用于去重检测")
 
 
 class DirectUploadCompleteResponse(BaseModel):
     storage_key: str
     ingest_job_id: str
+    duplicated: bool = Field(default=False, description="是否为重复素材（命中去重）")
+    asset_id: str = Field(default="", description="重复素材的 asset_id（duplicated=true 时返回）")
+
+
+class UploadAssetResponse(BaseModel):
+    storage_key: str
+    ingest_job_id: str
+    url: str = Field(..., description="Public URL of uploaded file")
+    duplicated: bool = Field(default=False, description="是否为重复素材（命中去重）")
+    asset_id: str = Field(default="", description="重复素材的 asset_id（duplicated=true 时返回）")

apps/web/e2e/core-generation.spec.ts

@@ -1,4 +1,4 @@
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "SmokePass123!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -20,7 +20,7 @@ const routeBrowserApiToTestApi = async (
 };
 
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,
@@ -40,6 +40,7 @@ async function loginWithRetry(
 
 type ProjectResponse = { id: string };
 type LibraryResponse = { id: string };
+type TemplateResponse = { id: string };
 type AssetListResponse = {
   items: Array<{
     id: string;
@@ -109,7 +110,6 @@ test.describe("Core generation flow", () => {
     expect(upload.status()).toBe(200);
 
     // Wait for asset to be ready
-    let sourceAssetId = "";
     await expect
       .poll(
         async () => {
@@ -121,13 +121,35 @@ test.describe("Core generation flow", () => {
           const data = (await assets.json()) as AssetListResponse;
           const asset = data.items.find((a) => a.name === sourceFileName);
           if (!asset) return "missing";
-          sourceAssetId = asset.id;
           return asset.status;
         },
         { timeout: 30_000, intervals: [1_000, 2_000, 3_000] },
       )
       .toBe("ready");
 
+    // Create an editing template so the generate page has at least one template
+    // (templates are now loaded from API; new users have none by default)
+    const template = await request.post(`${apiBase}/templates`, {
+      headers,
+      data: {
+        name: `E2E 测试模板 ${suffix}`,
+        mode: "pip",
+        estimated_duration: 30,
+        segments: [
+          {
+            segment_order: 1,
+            duration_min: 5,
+            duration_max: 30,
+            material_type: "video",
+          },
+        ],
+        tags: ["e2e"],
+      },
+    });
+    expect(template.status(), await template.text()).toBe(201);
+    const templateData = (await template.json()) as TemplateResponse;
+    expect(templateData.id).toBeTruthy();
+
     // Set auth in localStorage
     await page.addInitScript(
       ({ token, user }) => {
@@ -165,38 +187,30 @@ test.describe("Core generation flow", () => {
     await page.getByRole("button", { name: "下一步" }).click();
 
     // Step 2: select material
-    await expect(
-      page.getByRole("heading", { name: /选择素材/ }),
-    ).toBeVisible();
+    await expect(page.getByRole("heading", { name: /选择素材/ })).toBeVisible();
     const librarySelect = page.locator("select").first();
     await librarySelect.selectOption({ label: libraryName });
     const materialLabel = page.getByText(sourceFileName).locator("..");
-    await expect(
-      materialLabel.locator("input[type='checkbox']"),
-    ).toBeVisible({ timeout: 10_000 });
+    await expect(materialLabel.locator("input[type='checkbox']")).toBeVisible({
+      timeout: 10_000,
+    });
     await materialLabel.locator("input[type='checkbox']").check();
     await page.getByRole("button", { name: "下一步" }).click();
 
     // Step 3: title
-    await expect(
-      page.getByRole("heading", { name: /选择标题/ }),
-    ).toBeVisible();
+    await expect(page.getByRole("heading", { name: /选择标题/ })).toBeVisible();
     const titleText = `E2E Test ${suffix}`;
     await page.getByPlaceholder("输入自定义标题…").fill(titleText);
     await page.getByRole("button", { name: "下一步" }).click();
 
     // Step 4: voice
-    await expect(
-      page.getByRole("heading", { name: /选择配音/ }),
-    ).toBeVisible();
+    await expect(page.getByRole("heading", { name: /选择配音/ })).toBeVisible();
     const firstVoiceCard = page.locator(".xx-voice-choice-item").first();
     await firstVoiceCard.click();
     await page.getByRole("button", { name: "下一步" }).click();
 
     // Step 5: confirm and generate
-    await expect(
-      page.getByRole("heading", { name: /确认生成/ }),
-    ).toBeVisible();
+    await expect(page.getByRole("heading", { name: /确认生成/ })).toBeVisible();
 
     // Wait for plan creation API to be called
     const createPlanPromise = page.waitForResponse(
@@ -222,7 +236,7 @@ test.describe("Core generation flow", () => {
 
     // Generation may fail in test env (no worker), that's OK
     // Just verify the flow started - check page shows generation-related UI
-    const hasProgress = await page
+    await page
       .getByText(/生成中|生成完成|生成失败/)
       .isVisible({ timeout: 15_000 })
       .catch(() => false);

apps/web/e2e/core-titles.spec.ts

@@ -1,4 +1,4 @@
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "SmokePass123!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -21,7 +21,7 @@ const routeBrowserApiToTestApi = async (
 
 /** 登录操作，遇到 429 限流自动等待重试 */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,
@@ -60,6 +60,19 @@ test.describe("Title library flow", () => {
     const login = await loginWithRetry(request, email, PASSWORD);
     expect(login.status(), await login.text()).toBe(200);
     const loginData = (await login.json()) as { access_token: string };
+    const headers = { Authorization: `Bearer ${loginData.access_token}` };
+
+    // Create a title so the titles page has at least one title card to display
+    // (titles are loaded from API; new users have none by default)
+    const createTitle = await request.post(`${apiBase}/titles`, {
+      headers,
+      data: {
+        name: `E2E 测试标题 ${suffix}`,
+        text: `E2E 测试标题内容 ${suffix}`,
+        category: "default",
+      },
+    });
+    expect(createTitle.status(), await createTitle.text()).toBe(201);
 
     await page.addInitScript(
       ({ token, user }) => {

apps/web/e2e/core-upload.spec.ts

@@ -1,4 +1,4 @@
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "SmokePass123!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -21,7 +21,7 @@ const routeBrowserApiToTestApi = async (
 
 /** 登录操作，遇到 429 限流自动等待重试 */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,

apps/web/e2e/subscription.spec.ts

@@ -3,7 +3,7 @@
  *
  * 覆盖：路由守卫、订阅降级、过期处理、订阅状态检查
  */
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "Test123456!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -18,7 +18,7 @@ function uniqueUsername(prefix: string): string {
 
 /** 登录操作，遇到 429 限流自动等待重试 */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,
@@ -37,7 +37,7 @@ async function loginWithRetry(
 }
 
 /** 注册并登录，返回 { headers, email, username, userId } */
-async function createAuthedUser(request: any, label: string) {
+async function createAuthedUser(request: APIRequestContext, label: string) {
   const email = uniqueEmail(label);
   const username = uniqueUsername(label);
 

apps/web/e2e/test_asset.spec.ts

@@ -4,7 +4,7 @@
  * 覆盖：创建素材库、列出素材库、创建素材记录
  * 每个测试独立，先注册登录获取 auth token。
  */
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "Test123456!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -19,7 +19,7 @@ function uniqueUsername(prefix: string): string {
 
 /** 登录操作，遇到 429 限流自动等待重试 */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,
@@ -38,7 +38,7 @@ async function loginWithRetry(
 }
 
 /** 注册并登录，返回 { headers, email, username, userId } */
-async function createAuthedUser(request: any, label: string) {
+async function createAuthedUser(request: APIRequestContext, label: string) {
   const email = uniqueEmail(label);
   const username = uniqueUsername(label);
 
@@ -62,7 +62,7 @@ async function createAuthedUser(request: any, label: string) {
 
 /** 创建一个项目并返回 project id */
 async function createProject(
-  request: any,
+  request: APIRequestContext,
   headers: Record<string, string>,
   suffix: string,
 ): Promise<string> {
@@ -185,7 +185,7 @@ test.describe("素材库流程", () => {
     const items = data.items || [];
     expect(items.length, "应至少有 2 个素材库").toBeGreaterThanOrEqual(2);
 
-    const kinds = items.map((i: any) => i.kind);
+    const kinds = items.map((i: { kind: string }) => i.kind);
     expect(kinds).toContain("video");
     expect(kinds).toContain("image");
   });

apps/web/e2e/test_auth.spec.ts

@@ -4,7 +4,7 @@
  * 覆盖：注册（正向/反向）、登录（正向/反向）、登出、获取当前用户信息
  * 每个测试独立，使用随机邮箱避免冲突。
  */
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "Test123456!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -18,6 +18,7 @@ function uniqueUsername(prefix: string): string {
 }
 
 /** 从错误响应中提取错误消息文本，兼容新老格式 */
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
 function extractErrorMessage(body: any): string {
   if (!body) return "";
   // 新格式: { error: { code: "...", message: "..." } }
@@ -30,7 +31,7 @@ function extractErrorMessage(body: any): string {
 
 /** 登录操作，遇到 429 限流自动等待重试（最多等 65s） */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,

apps/web/e2e/test_project.spec.ts

@@ -4,7 +4,7 @@
  * 覆盖：创建项目、列出项目、获取项目详情
  * 每个测试独立，先注册登录获取 auth token。
  */
-import { expect, test } from "@playwright/test";
+import { expect, test, type APIRequestContext } from "@playwright/test";
 
 const PASSWORD = "Test123456!";
 const apiBase = process.env.E2E_API_BASE || "/api/v1";
@@ -19,7 +19,7 @@ function uniqueUsername(prefix: string): string {
 
 /** 登录操作，遇到 429 限流自动等待重试 */
 async function loginWithRetry(
-  request: any,
+  request: APIRequestContext,
   email: string,
   password: string,
   maxRetries = 2,
@@ -38,7 +38,7 @@ async function loginWithRetry(
 }
 
 /** 注册并登录，返回 { headers, email, username, userId } */
-async function createAuthedUser(request: any, label: string) {
+async function createAuthedUser(request: APIRequestContext, label: string) {
   const email = uniqueEmail(label);
   const username = uniqueUsername(label);
 

apps/web/src/api/assets.ts

@@ -19,6 +19,7 @@ export interface AssetItem {
   status?: string;
   classification_status?: string | null;
   quality_score?: number | null;
+  tag_ids?: string[];
   created_at?: string;
 }
 
@@ -114,6 +115,18 @@ export const createAssetLibrary = async (data: {
   return response.data;
 };
 
+/** 确保项目下指定 kind 的默认素材库存在（不存在则自动创建） */
+export const ensureDefaultLibrary = async (data: {
+  project_id: string;
+  kind: "video" | "voice" | "image";
+}): Promise<AssetLibraryItem> => {
+  const response = await apiClient.post(
+    "/asset-libraries/ensure-default",
+    data,
+  );
+  return response.data;
+};
+
 /** 删除素材库 */
 export const deleteAssetLibrary = async (libraryId: string): Promise<void> => {
   await apiClient.delete(`/asset-libraries/${libraryId}`);
@@ -129,6 +142,46 @@ export const getAssets = async (libraryId: string): Promise<AssetItem[]> => {
   return response.data.items || [];
 };
 
+/** 按类型获取素材（如 voice/video/image），支持可选筛选 */
+export const getAssetsByKind = async (
+  kind: string,
+  filters?: {
+    keyword?: string;
+    gender?: string;
+    style?: string;
+    tag_ids?: string[];
+  },
+): Promise<AssetItem[]> => {
+  const params: Record<string, string> = { kind };
+  if (filters?.keyword) params.keyword = filters.keyword;
+  if (filters?.gender) params.gender = filters.gender;
+  if (filters?.style) params.style = filters.style;
+  if (filters?.tag_ids?.length) params.tag_ids = filters.tag_ids.join(",");
+  const response = await apiClient.get("/assets", { params });
+  return response.data.items || [];
+};
+
+/** 创建素材（上传文件后调用，附带 metadata） */
+export const createAsset = async (data: {
+  library_id: string;
+  name: string;
+  storage_key: string;
+  mime_type: string;
+  metadata?: Record<string, unknown>;
+}): Promise<AssetItem> => {
+  const response = await apiClient.post("/assets", data);
+  return response.data;
+};
+
+/** 更新素材（名称、metadata 等） */
+export const updateAsset = async (
+  assetId: string,
+  data: { name?: string; metadata?: Record<string, unknown> },
+): Promise<AssetItem> => {
+  const response = await apiClient.put(`/assets/${assetId}`, data);
+  return response.data;
+};
+
 /** 更新素材审核状态 */
 export const updateAssetReviewStatus = async (
   assetId: string,
@@ -187,10 +240,11 @@ export const completeDirectUpload = async (data: {
   return response.data;
 };
 
-/** 直传上传（大文件推荐） */
+/** 直传上传（大文件推荐），支持可选进度回调 */
 export const uploadAssetDirect = async (data: {
   file: File;
   library_id: string;
+  onProgress?: (percent: number) => void;
 }): Promise<{ storage_key: string; ingest_job_id: string }> => {
   // 后端要求 project_id，前端自动获取默认项目
   const project = await getOrCreateDefaultProject();
@@ -209,13 +263,25 @@ export const uploadAssetDirect = async (data: {
   );
   directForm.append("file", data.file);
 
-  const uploadResponse = await fetch(prepared.upload_url, {
-    method: prepared.method,
-    body: directForm,
+  // 使用 XMLHttpRequest 以获取上传进度（fetch 不支持）
+  await new Promise<void>((resolve, reject) => {
+    const xhr = new XMLHttpRequest();
+    xhr.open(prepared.method, prepared.upload_url);
+    xhr.upload.onprogress = (e) => {
+      if (e.lengthComputable && data.onProgress) {
+        data.onProgress(Math.round((e.loaded / e.total) * 100));
+      }
+    };
+    xhr.onload = () => {
+      if (xhr.status >= 200 && xhr.status < 300) {
+        resolve();
+      } else {
+        reject(new Error(`OSS direct upload failed: ${xhr.status}`));
+      }
+    };
+    xhr.onerror = () => reject(new Error("OSS direct upload failed"));
+    xhr.send(directForm);
   });
-  if (!uploadResponse.ok) {
-    throw new Error(`OSS direct upload failed: ${uploadResponse.status}`);
-  }
 
   return completeDirectUpload({
     project_id: project.id,

apps/web/src/api/auth.ts

@@ -1,6 +1,7 @@
 /**
  * 认证相关 API
  */
+import axios from "axios";
 import apiClient from "./client";
 
 // 类型定义
@@ -71,6 +72,17 @@ export const login = async (data: LoginRequest): Promise<LoginResponse> => {
   return response.data;
 };
 
+// 刷新 access_token（使用裸 axios 避免拦截器递归）
+export const refreshAccessToken = async (
+  refreshToken: string,
+): Promise<LoginResponse> => {
+  const baseURL = apiClient.defaults.baseURL ?? "";
+  const response = await axios.post(`${baseURL}/auth/refresh`, {
+    refresh_token: refreshToken,
+  });
+  return response.data;
+};
+
 // 注册
 export const register = async (
   data: RegisterRequest,

apps/web/src/api/client.ts

@@ -5,6 +5,7 @@
 import axios, { AxiosError, InternalAxiosRequestConfig } from "axios";
 import { message } from "antd";
 import { useAuthStore } from "@/store/authStore";
+import { refreshAccessToken } from "./auth";
 
 // 创建 Axios 实例
 const apiClient = axios.create({
@@ -15,6 +16,24 @@ const apiClient = axios.create({
   },
 });
 
+// ── Token 刷新队列 ─────────────────────────────────────────
+let isRefreshing = false;
+let failedQueue: Array<{
+  resolve: (value?: unknown) => void;
+  reject: (reason?: unknown) => void;
+}> = [];
+
+const processQueue = (error: unknown | null, token: string | null = null) => {
+  failedQueue.forEach((prom) => {
+    if (error) {
+      prom.reject(error);
+    } else {
+      prom.resolve(token);
+    }
+  });
+  failedQueue = [];
+};
+
 // 请求拦截器：添加 Token
 apiClient.interceptors.request.use(
   (config: InternalAxiosRequestConfig) => {
@@ -29,15 +48,77 @@ apiClient.interceptors.request.use(
   },
 );
 
-// 响应拦截器：统一错误提示 + 处理未授权状态
+// 响应拦截器：统一错误提示 + 401 自动刷新 Token
 apiClient.interceptors.response.use(
   (response) => response,
   async (
     error: AxiosError<{ detail?: string; message?: string; msg?: string }>,
   ) => {
-    // 401 → 清除登录态
-    if (error.response?.status === 401) {
-      useAuthStore.getState().clearAuth();
+    const originalRequest = error.config as InternalAxiosRequestConfig & {
+      _retry?: boolean;
+    };
+
+    // 401 → 尝试刷新 Token
+    if (
+      error.response?.status === 401 &&
+      originalRequest &&
+      !originalRequest._retry
+    ) {
+      const refreshToken = useAuthStore.getState().refreshToken;
+
+      // 无 refresh_token → 直接登出
+      if (!refreshToken) {
+        useAuthStore.getState().clearAuth();
+        window.location.href = "/";
+        return Promise.reject(error);
+      }
+
+      // 已在刷新中 → 排队等待
+      if (isRefreshing) {
+        return new Promise((resolve, reject) => {
+          failedQueue.push({ resolve, reject });
+        }).then((token) => {
+          if (originalRequest.headers) {
+            originalRequest.headers.Authorization = `Bearer ${token}`;
+          }
+          return apiClient(originalRequest);
+        });
+      }
+
+      originalRequest._retry = true;
+      isRefreshing = true;
+
+      try {
+        const data = await refreshAccessToken(refreshToken);
+        const newAccessToken = data.access_token;
+        const newRefreshToken = data.refresh_token ?? refreshToken;
+
+        // 更新 Zustand + localStorage
+        useAuthStore
+          .getState()
+          .setAuth(
+            useAuthStore.getState().user!,
+            newAccessToken,
+            newRefreshToken,
+          );
+
+        // 处理排队的请求
+        processQueue(null, newAccessToken);
+
+        // 重试原始请求
+        if (originalRequest.headers) {
+          originalRequest.headers.Authorization = `Bearer ${newAccessToken}`;
+        }
+        return apiClient(originalRequest);
+      } catch (refreshError) {
+        // 刷新失败 → 登出
+        processQueue(refreshError, null);
+        useAuthStore.getState().clearAuth();
+        window.location.href = "/";
+        return Promise.reject(refreshError);
+      } finally {
+        isRefreshing = false;
+      }
     }
 
     // 提取后端返回的错误信息（detail / message / msg）

apps/web/src/api/tags.ts

@@ -0,0 +1,46 @@
+/**
+ * 标签 CRUD API
+ * P3 标签体系：对接后端标签表
+ */
+import apiClient from "./client";
+
+export interface TagItem {
+  id: string;
+  name: string;
+  created_at?: string;
+  usage_count?: number;
+}
+
+/** 获取当前用户所有标签 */
+export const getTags = async (): Promise<TagItem[]> => {
+  const response = await apiClient.get("/tags");
+  return response.data.items || [];
+};
+
+/** 创建标签（同名返回 409） */
+export const createTag = async (name: string): Promise<TagItem> => {
+  const response = await apiClient.post("/tags", { name });
+  return response.data;
+};
+
+/** 删除标签（同时清理素材关联） */
+export const deleteTag = async (tagId: string): Promise<void> => {
+  await apiClient.delete(`/tags/${tagId}`);
+};
+
+/** 为素材添加标签（最多 50 个） */
+export const tagAsset = async (
+  assetId: string,
+  tagIds: string[],
+): Promise<void> => {
+  if (tagIds.length === 0) return;
+  await apiClient.post(`/assets/${assetId}/tags`, { tag_ids: tagIds });
+};
+
+/** 移除素材的某个标签 */
+export const untagAsset = async (
+  assetId: string,
+  tagId: string,
+): Promise<void> => {
+  await apiClient.delete(`/assets/${assetId}/tags/${tagId}`);
+};

apps/web/src/api/tts.ts

@@ -122,6 +122,20 @@ export const getTTSJobs = async (
   return response.data;
 };
 
+/** 存为素材请求参数 */
+export interface SaveTtsToLibraryRequest {
+  name?: string;
+  tag_ids?: string[];
+}
+
+/** 将 TTS 合成结果保存到配音素材库 */
+export const saveTtsToLibrary = async (
+  jobId: string,
+  data?: SaveTtsToLibraryRequest,
+): Promise<void> => {
+  await apiClient.post(`/tts/jobs/${jobId}/save-to-library`, data ?? {});
+};
+
 /** 删除 TTS 任务 */
 export const deleteTTSJob = async (jobId: string): Promise<void> => {
   await apiClient.delete(`/tts/jobs/${jobId}`);

apps/web/src/components/layout/MainLayout.tsx

@@ -20,6 +20,7 @@ export interface SidebarContextValue {
   collapsed: boolean;
 }
 
+// eslint-disable-next-line react-refresh/only-export-components
 export const SidebarContext = React.createContext<SidebarContextValue>({
   collapsed: false,
 });

apps/web/src/components/layout/PageHead.tsx

@@ -62,6 +62,7 @@ const ROUTE_TITLE_MAP: Record<string, string> = {
   "/app/editing-planner": "剪辑规划",
   "/app/my-templates": "我的模板",
   "/app/voice-clone": "我的音色",
+  "/app/voice-materials": "配音素材库",
   "/app/accounts": "账号管理",
   "/app/duplication": "查重",
   "/app/duplication/results": "查重结果",

apps/web/src/components/modals/CloneVoiceModal.tsx

@@ -10,6 +10,7 @@ import React, { useState, useCallback, useRef } from "react";
 import { Modal, Button } from "@/components/ui";
 import { createVoiceClone, toVoiceClone } from "@/api/voiceClone";
 import type { VoiceClone } from "@/api/voiceClone";
+import { uploadAsset } from "@/api/assets";
 import "./clone-voice-modal.css";
 
 /* ── 类型定义 ───────────────────────────────────────────── */
@@ -121,12 +122,24 @@ const CloneVoiceModal: React.FC<CloneVoiceModalProps> = ({
     setStep("uploading");
 
     try {
-      // Mock：模拟上传 + 克隆过程
+      // 先上传音频文件获取真实 URL
+      let audioUrl: string;
+      if (selectedFile) {
+        const formData = new FormData();
+        formData.append("file", selectedFile);
+        formData.append("kind", "voice");
+        const uploadResult = await uploadAsset(formData);
+        audioUrl = uploadResult.url;
+      } else {
+        // 录制功能暂未实现，提示用户上传
+        setStep("input");
+        return;
+      }
+
+      // 提交克隆请求
       const result = await createVoiceClone({
         name,
-        audio_url: selectedFile
-          ? `mock://${selectedFile.name}`
-          : "mock://recorded-audio",
+        audio_url: audioUrl,
       });
 
       setStep("success");

apps/web/src/components/ui/Input.tsx

@@ -2,6 +2,7 @@
  * V21 Input 输入框
  * 封装 Ant Design Input，应用 V21 设计系统样式
  */
+/* eslint-disable react-refresh/only-export-components */
 import React from "react";
 import { Input as AntInput } from "antd";
 import type { InputProps as AntInputProps } from "antd";

apps/web/src/config/navigation.tsx

@@ -67,6 +67,12 @@ export const NAV_ITEMS: NavItem[] = [
     path: "/app/voice-clone",
     icon: <AudioOutlined />,
   },
+  {
+    key: "voice-materials",
+    label: "配音素材库",
+    path: "/app/voice-materials",
+    icon: <AudioOutlined />,
+  },
   {
     key: "templates",
     label: "模板库",
@@ -153,6 +159,12 @@ export const NAV_GROUPS: NavGroup[] = [
         path: "/app/voice-clone",
         icon: <AudioOutlined />,
       },
+      {
+        key: "voice-materials",
+        label: "配音素材库",
+        path: "/app/voice-materials",
+        icon: <AudioOutlined />,
+      },
       {
         key: "titles",
         label: "标题库",

apps/web/src/hooks/useAuth.ts

@@ -75,7 +75,7 @@ export const useLogout = () => {
     } finally {
       clearAuth();
       queryClient.clear();
-      navigate("/login");
+      navigate("/");
     }
   };
 

apps/web/src/index.css

@@ -25,6 +25,15 @@ body {
   background-color: var(--bg-secondary);
 }
 
+/* ── 自定义字体 ── */
+@font-face {
+  font-family: "华康俪金黑";
+  src: url("/fonts/DFLiJinHei-W8.ttf") format("truetype");
+  font-weight: 700;
+  font-style: normal;
+  font-display: swap;
+}
+
 /* 滚动条 - V21 样式 */
 ::-webkit-scrollbar {
   width: 8px;

apps/web/src/pages/assets/AssetLibrary.tsx

@@ -10,7 +10,6 @@ import {
   SearchOutlined,
   InboxOutlined,
   VideoCameraOutlined,
-  SoundOutlined,
   PictureOutlined,
   PlayCircleOutlined,
   CheckOutlined,
@@ -37,7 +36,7 @@ import "./assets.css";
 /* ============================================================
  * 类型
  * ============================================================ */
-type AssetKind = "video" | "voice" | "image";
+type AssetKind = "video" | "image";
 type StatusType = "ok" | "warn" | "bad" | "info";
 
 interface LibraryItem {
@@ -67,7 +66,6 @@ interface AssetItem {
 /** 根据 mime_type 推断前端 AssetKind */
 const inferKind = (mimeType: string): AssetKind => {
   if (mimeType.startsWith("video/")) return "video";
-  if (mimeType.startsWith("audio/")) return "voice";
   return "image";
 };
 
@@ -99,7 +97,7 @@ const formatDuration = (seconds: number): string => {
 const mapLibrary = (item: AssetLibraryItem): LibraryItem => ({
   id: item.id,
   name: item.name,
-  kind: item.kind || inferKind("video"),
+  kind: (item.kind === "voice" ? "video" : item.kind) || inferKind("video"),
   count: item.asset_count ?? 0,
 });
 
@@ -110,14 +108,16 @@ const mapAsset = (item: ApiAssetItem): AssetItem => {
     item.classification_status ?? undefined,
   );
   const metadata = item.metadata || {};
+  const kind = inferKind(item.mime_type || "");
   return {
     id: item.id,
     name: item.name,
-    kind: inferKind(item.mime_type || ""),
+    kind,
+    // 视频类型不能用 file_url 做缩略图（是视频文件，<img> 无法渲染）
     thumbUrl:
       (item.thumbnail_url as string | undefined) ||
-      (item.file_url as string | undefined) ||
-      (metadata.thumbnail_url as string | undefined),
+      (metadata.thumbnail_url as string | undefined) ||
+      (kind !== "video" ? (item.file_url as string | undefined) : undefined),
     fileUrl:
       (item.file_url as string | undefined) ||
       (metadata.file_url as string | undefined),
@@ -147,8 +147,6 @@ const kindIcon = (kind: AssetKind) => {
   switch (kind) {
     case "video":
       return <VideoCameraOutlined />;
-    case "voice":
-      return <SoundOutlined />;
     case "image":
       return <PictureOutlined />;
   }
@@ -158,8 +156,6 @@ const kindLabel = (kind: AssetKind) => {
   switch (kind) {
     case "video":
       return "视频";
-    case "voice":
-      return "配音";
     case "image":
       return "图片";
   }
@@ -170,8 +166,6 @@ const thumbGradient = (kind: AssetKind): string => {
   switch (kind) {
     case "video":
       return "linear-gradient(135deg, #312e81 0%, #4f46e5 50%, #6366f1 100%)";
-    case "voice":
-      return "linear-gradient(135deg, #064e3b 0%, #059669 50%, #10b981 100%)";
     case "image":
       return "linear-gradient(135deg, #78350f 0%, #d97706 50%, #f59e0b 100%)";
   }
@@ -244,7 +238,7 @@ const AssetCard: React.FC<{
       )}
 
       {/* 视频/配音类显示播放按钮 */}
-      {(asset.kind === "video" || asset.kind === "voice") && (
+      {asset.kind === "video" && (
         <span
           className="xx-asset-play"
           onClick={(e) => {
@@ -361,6 +355,7 @@ const AssetLibrary: React.FC = () => {
 
   /* 上传 */
   const [uploading, setUploading] = useState(false);
+  const [uploadProgress, setUploadProgress] = useState(0);
 
   /* 新建素材库 */
   const [createModalOpen, setCreateModalOpen] = useState(false);
@@ -434,18 +429,25 @@ const AssetLibrary: React.FC = () => {
     }
 
     setUploading(true);
+    setUploadProgress(0);
     try {
       if (file.size > LARGE_FILE_THRESHOLD) {
         message.info(`大文件 "${file.name}" 将使用直传上传`);
       }
-      await uploadAssetDirect({ file, library_id: effectiveLibId });
+      await uploadAssetDirect({
+        file,
+        library_id: effectiveLibId,
+        onProgress: (pct) => setUploadProgress(pct),
+      });
       message.success(`"${file.name}" 上传成功`);
       queryClient.invalidateQueries({ queryKey: ["assets"] });
       queryClient.invalidateQueries({ queryKey: ["asset-libraries"] });
-    } catch {
-      message.error(`"${file.name}" 上传失败`);
+    } catch (err: unknown) {
+      const detail = err instanceof Error ? err.message : "";
+      message.error(`"${file.name}" 上传失败${detail ? `：${detail}` : ""}`);
     } finally {
       setUploading(false);
+      setUploadProgress(0);
     }
     return false;
   };
@@ -532,6 +534,54 @@ const AssetLibrary: React.FC = () => {
 
   return (
     <div className="xx-assets-page">
+      {/* ─── 上传进度弹窗（圆形动画 + 百分比） ─── */}
+      <AntModal
+        open={uploading}
+        footer={null}
+        closable={false}
+        centered
+        width={260}
+        maskClosable={false}
+        className="xx-upload-progress-modal"
+      >
+        <div className="xx-upload-progress-body">
+          <svg
+            className="xx-upload-progress-ring"
+            viewBox="0 0 120 120"
+            width={120}
+            height={120}
+          >
+            {/* 背景圆环 */}
+            <circle
+              cx="60"
+              cy="60"
+              r="52"
+              fill="none"
+              stroke="var(--border-primary, #e5e7eb)"
+              strokeWidth="8"
+            />
+            {/* 进度圆弧 */}
+            <circle
+              cx="60"
+              cy="60"
+              r="52"
+              fill="none"
+              stroke="var(--primary-color, #6366f1)"
+              strokeWidth="8"
+              strokeLinecap="round"
+              strokeDasharray={`${2 * Math.PI * 52}`}
+              strokeDashoffset={`${2 * Math.PI * 52 * (1 - uploadProgress / 100)}`}
+              transform="rotate(-90 60 60)"
+              style={{ transition: "stroke-dashoffset 0.3s ease" }}
+            />
+          </svg>
+          <div className="xx-upload-progress-text">
+            <span className="xx-upload-progress-pct">{uploadProgress}%</span>
+            <span className="xx-upload-progress-label">上传中…</span>
+          </div>
+        </div>
+      </AntModal>
+
       {/* 两栏布局 */}
       <div className="xx-assets-layout">
         {/* ─── 左侧：素材库列表 ─── */}
@@ -588,7 +638,7 @@ const AssetLibrary: React.FC = () => {
             beforeUpload={handleUpload}
             showUploadList={false}
             multiple
-            accept="video/*,audio/*,image/*"
+            accept="video/*,image/*"
           >
             <div className="xx-asset-upload-zone">
               <p className="xx-asset-upload-icon">
@@ -598,7 +648,7 @@ const AssetLibrary: React.FC = () => {
                 {uploading ? "上传中..." : "点击或拖拽文件到此区域上传"}
               </p>
               <p className="xx-asset-upload-hint">
-                支持视频、音频、图片，单文件不超过 2GB
+                支持视频、图片，单文件不超过 2GB
               </p>
             </div>
           </Upload.Dragger>
@@ -621,7 +671,6 @@ const AssetLibrary: React.FC = () => {
                 options={[
                   { value: "all", label: "全部类型" },
                   { value: "video", label: "视频" },
-                  { value: "voice", label: "配音" },
                   { value: "image", label: "图片" },
                 ]}
               />
@@ -752,7 +801,6 @@ const AssetLibrary: React.FC = () => {
               style={{ width: "100%" }}
               options={[
                 { value: "video", label: "视频" },
-                { value: "voice", label: "配音" },
                 { value: "image", label: "图片" },
               ]}
             />

apps/web/src/pages/assets/assets.css

@@ -587,3 +587,40 @@
     grid-template-columns: 1fr;
   }
 }
+
+/* ─── 上传进度弹窗 ─── */
+.xx-upload-progress-modal .ant-modal-content {
+  padding: 24px 16px 20px;
+  border-radius: 16px;
+}
+
+.xx-upload-progress-body {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: 16px;
+  padding: 8px 0;
+}
+
+.xx-upload-progress-ring {
+  display: block;
+}
+
+.xx-upload-progress-text {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: 4px;
+}
+
+.xx-upload-progress-pct {
+  font-size: 22px;
+  font-weight: 700;
+  color: var(--primary-color, #6366f1);
+  line-height: 1;
+}
+
+.xx-upload-progress-label {
+  font-size: 13px;
+  color: var(--text-secondary, #6b7280);
+}

apps/web/src/pages/auth/Login.css

@@ -31,7 +31,11 @@
   font-size: 15px;
   font-weight: 700;
   border-radius: var(--radius-sm);
-  background: linear-gradient(135deg, var(--color-primary-500), var(--color-primary-600));
+  background: linear-gradient(
+    135deg,
+    var(--color-primary-500),
+    var(--color-primary-600)
+  );
   border: none;
   box-shadow: 0 4px 14px
     color-mix(in srgb, var(--primary-color) 30%, transparent);

apps/web/src/pages/auth/Register.css

@@ -31,7 +31,11 @@
   font-size: 15px;
   font-weight: 700;
   border-radius: var(--radius-sm);
-  background: linear-gradient(135deg, var(--color-primary-500), var(--color-primary-600));
+  background: linear-gradient(
+    135deg,
+    var(--color-primary-500),
+    var(--color-primary-600)
+  );
   border: none;
   box-shadow: 0 4px 14px
     color-mix(in srgb, var(--primary-color) 30%, transparent);

apps/web/src/pages/editing-planner/EditingPlanner.css

@@ -347,6 +347,53 @@
   background: var(--primary-soft, #eef2ff);
 }
 
+/* 左侧 Tab 切换 */
+.ep-left-tabs {
+  display: flex;
+  border-bottom: 1px solid var(--border-color, #e2e8f0);
+  background: var(--bg-secondary, #f8fafc);
+}
+
+.ep-left-tab {
+  flex: 1;
+  padding: 12px 16px;
+  font-size: 13px;
+  font-weight: 500;
+  color: var(--text-secondary, #64748b);
+  background: transparent;
+  border: none;
+  border-bottom: 2px solid transparent;
+  cursor: pointer;
+  transition: all 0.2s ease;
+}
+
+.ep-left-tab:hover {
+  color: var(--text-primary, #1e293b);
+  background: var(--bg-primary, #ffffff);
+}
+
+.ep-left-tab.active {
+  color: var(--primary, #6366f1);
+  border-bottom-color: var(--primary, #6366f1);
+  background: var(--bg-primary, #ffffff);
+}
+
+/* 素材 Tab 容器 */
+.ep-assets-tab {
+  flex: 1;
+  overflow-y: auto;
+  padding: 12px;
+}
+
+.ep-assets-tab::-webkit-scrollbar {
+  width: 6px;
+}
+
+.ep-assets-tab::-webkit-scrollbar-thumb {
+  background: var(--border-color, #e2e8f0);
+  border-radius: 3px;
+}
+
 /* 模板列表 */
 .ep-template-list {
   flex: 1;
@@ -504,7 +551,11 @@
   display: flex;
   align-items: center;
   justify-content: center;
-  background: linear-gradient(135deg, var(--ep-bg-card-hover, #1a1a2e), var(--ep-bg-deepest, #0a0a14));
+  background: linear-gradient(
+    135deg,
+    var(--ep-bg-card-hover, #1a1a2e),
+    var(--ep-bg-deepest, #0a0a14)
+  );
   height: calc(100% - 20px);
   position: relative;
 }
@@ -554,6 +605,34 @@
   color: var(--ep-text-secondary, #9ca3af);
 }
 
+/* 标题/字幕实时预览叠加层 */
+.ep-preview-title,
+.ep-preview-subtitle {
+  position: absolute;
+  left: 4px;
+  right: 4px;
+  text-align: center;
+  color: #fff;
+  line-height: 1.3;
+  pointer-events: none;
+  z-index: 5;
+  word-break: break-all;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.ep-preview-title {
+  text-shadow: 1px 1px 3px rgba(0, 0, 0, 0.7);
+}
+
+.ep-preview-subtitle {
+  background: rgba(0, 0, 0, 0.55);
+  padding: 2px 4px;
+  border-radius: 3px;
+  text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.5);
+}
+
 /* 封面预览 */
 .ep-cover-preview {
   width: 150px;
@@ -572,7 +651,11 @@
 .ep-cover-image {
   width: 100%;
   height: 100%;
-  background: linear-gradient(135deg, var(--ep-bg-card-hover, #1a1a2e), var(--ep-bg-card, #13131f));
+  background: linear-gradient(
+    135deg,
+    var(--ep-bg-card-hover, #1a1a2e),
+    var(--ep-bg-card, #13131f)
+  );
   display: flex;
   align-items: center;
   justify-content: center;
@@ -764,8 +847,8 @@
   flex: 1;
   display: flex;
   align-items: center;
-  gap: 12px;
-  padding: 12px 16px;
+  gap: 8px;
+  padding: 6px 16px;
   overflow-x: auto;
   overflow-y: hidden;
   background: var(--bg-primary, #ffffff);
@@ -781,11 +864,11 @@
   border-radius: 3px;
 }
 
-/* 水平片段卡片 — 统一高度96px、圆角、hover高亮 */
+/* 水平片段卡片 — 统一高度64px、圆角、hover高亮 */
 .ep-clip-card {
   width: 100px;
   min-width: 100px;
-  height: 96px;
+  height: 64px;
   background: var(--bg-primary, #ffffff);
   border: 1px solid var(--border-color, #e2e2e2);
   border-radius: var(--radius-sm, 14px);
@@ -805,7 +888,9 @@
 
 .ep-clip-card.selected {
   border-color: var(--primary, #6366f1);
-  box-shadow: 0 0 0 2px var(--primary-soft, #eef2ff), 0 4px 12px rgba(99, 102, 241, 0.2);
+  box-shadow:
+    0 0 0 2px var(--primary-soft, #eef2ff),
+    0 4px 12px rgba(99, 102, 241, 0.2);
 }
 
 .ep-clip-card.drag-over {
@@ -818,8 +903,12 @@
 }
 
 .ep-clip-thumbnail {
-  height: 52px;
-  background: linear-gradient(135deg, var(--bg-secondary, #f8fafc), var(--border-light, #f1f5f9));
+  height: 34px;
+  background: linear-gradient(
+    135deg,
+    var(--bg-secondary, #f8fafc),
+    var(--border-light, #f1f5f9)
+  );
   display: flex;
   align-items: center;
   justify-content: center;
@@ -828,7 +917,7 @@
 }
 
 .ep-clip-info {
-  padding: 6px 8px;
+  padding: 4px 6px;
   flex: 1;
   display: flex;
   flex-direction: column;
@@ -898,11 +987,243 @@
 .ep-track-empty {
   flex: 1;
   display: flex;
+  flex-direction: column;
   align-items: center;
   justify-content: center;
+  gap: 8px;
   color: var(--text-tertiary, #94a3b8);
   font-size: 12px;
-  min-height: 96px;
+  min-height: 64px;
+}
+
+.ep-track-empty-icon {
+  font-size: 28px;
+  margin-bottom: 2px;
+}
+
+.ep-track-empty-text {
+  color: var(--text-tertiary, #94a3b8);
+  font-size: 12px;
+}
+
+.ep-track-empty-add-btn {
+  margin-top: 4px;
+  padding: 6px 18px;
+  background: var(--primary, #6366f1);
+  color: #fff;
+  border: none;
+  border-radius: var(--radius-sm, 14px);
+  font-size: 13px;
+  font-weight: 500;
+  cursor: pointer;
+  transition: all var(--transition-fast, 0.15s ease);
+}
+
+.ep-track-empty-add-btn:hover {
+  background: var(--primary-hover, #4f46e5);
+  transform: translateY(-1px);
+  box-shadow: 0 2px 8px rgba(99, 102, 241, 0.3);
+}
+
+/* 轨道末尾 "+" 添加卡片 — 对齐 V21 原型虚线 "+" 卡 */
+.ep-track-add-card-wrapper {
+  position: relative;
+  flex-shrink: 0;
+}
+
+.ep-track-add-card {
+  min-width: 60px;
+  height: 64px;
+  border: 2px dashed var(--line, #e2e8f0);
+  border-radius: 12px;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  cursor: pointer;
+  font-size: 24px;
+  color: #94a3b8;
+  transition: all 0.2s ease;
+  user-select: none;
+}
+
+.ep-track-add-card:hover {
+  border-color: var(--primary, #6366f1);
+  color: var(--primary, #6366f1);
+  background: var(--primary-soft, #eef2ff);
+}
+
+.ep-track-add-card.locked {
+  cursor: not-allowed;
+  opacity: 0.5;
+  font-size: 18px;
+}
+
+.ep-track-add-card.locked:hover {
+  border-color: var(--line, #e2e8f0);
+  color: #94a3b8;
+  background: none;
+}
+
+/* picker 已改为 Portal 渲染，不再需要相对定位 */
+
+/* 添加片段按钮 + 选择器 */
+.ep-add-clip-wrapper {
+  position: relative;
+  display: inline-block;
+}
+
+.ep-add-clip-btn {
+  padding: 4px 12px;
+  background: var(--primary, #6366f1);
+  color: #fff;
+  border: none;
+  border-radius: var(--radius-sm, 14px);
+  font-size: 12px;
+  font-weight: 500;
+  cursor: pointer;
+  transition: all var(--transition-fast, 0.15s ease);
+  white-space: nowrap;
+}
+
+.ep-add-clip-btn:hover {
+  background: var(--primary-hover, #4f46e5);
+  transform: translateY(-1px);
+  box-shadow: 0 2px 8px rgba(99, 102, 241, 0.3);
+}
+
+/* ── "+" 卡片类型+时长选择器 ── */
+.ep-add-clip-picker {
+  position: absolute;
+  top: calc(100% + 6px);
+  right: 0;
+  width: 180px;
+  background: var(--bg-primary, #ffffff);
+  border: 1px solid var(--line, #e2e8f0);
+  border-radius: var(--radius-md, 18px);
+  box-shadow: 0 8px 24px rgba(0, 0, 0, 0.12);
+  z-index: 100;
+  padding: 10px;
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+}
+
+/* fixed 定位版本 — 不受任何父容器 overflow 裁剪 */
+.ep-add-clip-picker--portal {
+  position: fixed !important;
+  z-index: 1000;
+}
+
+.ep-add-clip-picker-title {
+  font-size: 13px;
+  font-weight: 600;
+  color: var(--text-primary, #1e293b);
+  padding-bottom: 6px;
+  border-bottom: 1px solid var(--line, #e2e8f0);
+}
+
+.ep-add-clip-type-row,
+.ep-add-clip-duration-row {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.ep-add-clip-type-label {
+  font-size: 12px;
+  color: var(--text-secondary, #64748b);
+  font-weight: 500;
+  flex-shrink: 0;
+  min-width: 36px;
+}
+
+.ep-add-clip-type-btn {
+  padding: 5px 10px;
+  border-radius: var(--radius-xs, 8px);
+  border: 1px solid var(--line, #e2e8f0);
+  background: var(--bg-primary, #ffffff);
+  color: var(--text-secondary, #64748b);
+  font-size: 11px;
+  cursor: pointer;
+  transition: all 0.15s ease;
+  display: flex;
+  align-items: center;
+  gap: 4px;
+  white-space: nowrap;
+}
+
+.ep-add-clip-type-btn:hover {
+  border-color: var(--primary, #6366f1);
+  color: var(--primary, #6366f1);
+}
+
+.ep-add-clip-type-btn.active {
+  border-color: var(--primary, #6366f1);
+  background: var(--primary-soft, #eef2ff);
+  color: var(--primary, #6366f1);
+  font-weight: 500;
+}
+
+.ep-add-clip-duration-unit {
+  font-size: 12px;
+  color: var(--text-tertiary, #94a3b8);
+}
+
+.ep-add-clip-confirm-btn {
+  width: 100%;
+  height: 30px;
+  background: var(--primary, #6366f1);
+  color: var(--text-inverse, #ffffff);
+  border: none;
+  border-radius: var(--radius-xs, 8px);
+  font-size: 12px;
+  font-weight: 500;
+  cursor: pointer;
+  transition: all 0.15s ease;
+  margin-top: 4px;
+}
+
+.ep-add-clip-confirm-btn:hover {
+  background: var(--primary-hover, #4f46e5);
+  transform: translateY(-1px);
+  box-shadow: 0 2px 8px rgba(99, 102, 241, 0.3);
+}
+
+/* ── 片段详情 — 类型选择器 ── */
+.ep-clip-type-selector {
+  display: flex;
+  gap: 6px;
+}
+
+.ep-clip-type-btn {
+  padding: 5px 12px;
+  border-radius: var(--radius-xs, 8px);
+  border: 1px solid var(--line, #e2e8f0);
+  background: var(--bg-primary, #ffffff);
+  color: var(--text-secondary, #64748b);
+  font-size: 11px;
+  cursor: pointer;
+  transition: all 0.15s ease;
+  display: flex;
+  align-items: center;
+  gap: 4px;
+}
+
+.ep-clip-type-btn:hover:not(.disabled) {
+  border-color: var(--primary, #6366f1);
+  color: var(--primary, #6366f1);
+}
+
+.ep-clip-type-btn.active {
+  border-color: var(--primary, #6366f1);
+  background: var(--primary-soft, #eef2ff);
+  color: var(--primary, #6366f1);
+  font-weight: 500;
+}
+
+.ep-clip-type-btn.disabled {
+  opacity: 0.4;
+  cursor: not-allowed;
 }
 
 /* 一镜到底提示 */
@@ -1119,6 +1440,61 @@
   color: var(--primary, #6366f1);
 }
 
+/* ── 标题预设样式 — 剪映风格方形网格，示例文字"标题"预览完整效果 ── */
+.ep-title-presets-grid {
+  display: grid;
+  grid-template-columns: repeat(4, 1fr);
+  gap: 8px;
+}
+
+.ep-title-preset-card {
+  aspect-ratio: 1;
+  background: #ffffff;
+  border: 2px solid var(--border-color, #e2e8f0);
+  border-radius: var(--radius-xs, 8px);
+  cursor: pointer;
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  justify-content: center;
+  gap: 4px;
+  padding: 6px 4px;
+  transition: all 0.15s ease;
+  overflow: hidden;
+}
+
+.ep-title-preset-card:hover {
+  border-color: var(--primary, #6366f1);
+  background: #f8fafc;
+  transform: translateY(-1px);
+}
+
+.ep-title-preset-card.active {
+  border-color: var(--primary, #6366f1);
+  background: #eef2ff;
+  box-shadow: 0 0 0 2px rgba(99, 102, 241, 0.3);
+}
+
+.ep-title-preset-preview-text {
+  line-height: 1.2;
+  text-align: center;
+  word-break: keep-all;
+  white-space: nowrap;
+}
+
+.ep-title-preset-card-label {
+  font-size: 9px;
+  color: #64748b;
+  line-height: 1;
+  text-align: center;
+  margin-top: auto;
+}
+
+.ep-title-preset-card.active .ep-title-preset-card-label {
+  color: var(--primary, #6366f1);
+  font-weight: 500;
+}
+
 /* AI推荐按钮 */
 .ep-ai-recommend-btn {
   width: 100%;
@@ -1333,7 +1709,9 @@
   stroke: var(--primary, #6366f1);
   stroke-width: 8;
   stroke-linecap: round;
-  transition: stroke-dashoffset 0.5s ease, stroke 0.3s;
+  transition:
+    stroke-dashoffset 0.5s ease,
+    stroke 0.3s;
 }
 
 .ep-gen-progress-pct {
@@ -1475,7 +1853,9 @@
   font-size: 11px;
   padding: 2px 4px;
   border-radius: 3px;
-  transition: color 0.2s, background 0.2s;
+  transition:
+    color 0.2s,
+    background 0.2s;
   display: inline-flex;
   align-items: center;
   gap: 3px;
@@ -1490,11 +1870,15 @@
   color: var(--primary-dark);
 }
 
-
 /* Skeleton loading */
 @keyframes ep-skeleton-pulse {
-  0%, 100% { opacity: 0.4; }
-  50% { opacity: 0.8; }
+  0%,
+  100% {
+    opacity: 0.4;
+  }
+  50% {
+    opacity: 0.8;
+  }
 }
 
 .ep-skeleton {
@@ -1551,6 +1935,128 @@
   flex-shrink: 0;
 }
 
+/* ═══ 配音素材选择器 ═══ */
+.ep-clip-detail-select {
+  width: 100%;
+  padding: 6px 8px;
+  font-size: 12px;
+  color: var(--text-primary, #e2e8f0);
+  background: var(--bg-tertiary, #1e293b);
+  border: 1px solid var(--border-color, #334155);
+  border-radius: 6px;
+  outline: none;
+  cursor: pointer;
+  transition: border-color 0.15s;
+}
+
+.ep-clip-detail-select:hover {
+  border-color: var(--primary, #6366f1);
+}
+
+.ep-clip-detail-select:focus {
+  border-color: var(--primary, #6366f1);
+  box-shadow: 0 0 0 2px rgba(99, 102, 241, 0.15);
+}
+
+.ep-voice-upload-btn {
+  width: 100%;
+  margin-top: 6px;
+  padding: 6px 0;
+  font-size: 12px;
+  color: var(--primary, #6366f1);
+  background: transparent;
+  border: 1px dashed var(--primary, #6366f1);
+  border-radius: 6px;
+  cursor: pointer;
+  transition:
+    background 0.15s,
+    color 0.15s;
+}
+
+.ep-voice-upload-btn:hover {
+  background: rgba(99, 102, 241, 0.08);
+}
+
+/* 配音标签行（含刷新按钮） */
+.ep-clip-detail-label {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+}
+
+.ep-voice-refresh-btn {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  width: 20px;
+  height: 20px;
+  padding: 0;
+  font-size: 11px;
+  line-height: 1;
+  color: var(--text-secondary, #94a3b8);
+  background: transparent;
+  border: none;
+  border-radius: 4px;
+  cursor: pointer;
+  transition:
+    background 0.15s,
+    color 0.15s;
+}
+
+.ep-voice-refresh-btn:hover {
+  color: var(--primary, #6366f1);
+  background: rgba(99, 102, 241, 0.1);
+}
+
+.ep-voice-refresh-btn:disabled {
+  opacity: 0.5;
+  cursor: not-allowed;
+}
+
+/* 配音选择行（select + 试听按钮） */
+.ep-voice-select-row {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+}
+
+.ep-voice-select-row .ep-clip-detail-select {
+  flex: 1;
+  min-width: 0;
+}
+
+.ep-voice-preview-btn {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  width: 28px;
+  height: 28px;
+  flex-shrink: 0;
+  font-size: 13px;
+  line-height: 1;
+  color: var(--text-primary, #e2e8f0);
+  background: var(--bg-tertiary, #1e293b);
+  border: 1px solid var(--border-color, #334155);
+  border-radius: 6px;
+  cursor: pointer;
+  transition:
+    background 0.15s,
+    border-color 0.15s;
+}
+
+.ep-voice-preview-btn:hover {
+  background: var(--bg-hover, #2d3a4f);
+  border-color: var(--primary, #6366f1);
+}
+
+.ep-voice-loading,
+.ep-voice-empty {
+  padding: 8px 0;
+  font-size: 12px;
+  color: var(--text-secondary, #94a3b8);
+  text-align: center;
+}
+
 /* ═══ 预览区 — 补充样式 ═══ */
 .ep-phone-empty-hint {
   color: rgba(255, 255, 255, 0.5);
@@ -1654,8 +2160,12 @@
 }
 
 @keyframes ep-fade-in {
-  from { opacity: 0; }
-  to { opacity: 1; }
+  from {
+    opacity: 0;
+  }
+  to {
+    opacity: 1;
+  }
 }
 
 .ep-modal {
@@ -1668,8 +2178,14 @@
 }
 
 @keyframes ep-scale-in {
-  from { opacity: 0; transform: scale(0.95); }
-  to { opacity: 1; transform: scale(1); }
+  from {
+    opacity: 0;
+    transform: scale(0.95);
+  }
+  to {
+    opacity: 1;
+    transform: scale(1);
+  }
 }
 
 .ep-modal-header {
@@ -1751,4 +2267,3 @@
   text-align: center;
   min-height: 80px;
 }
-

apps/web/src/pages/editing-planner/EditingPlanner.tsx

@@ -4,13 +4,8 @@
  */
 import React, { useState, useCallback, useEffect } from "react";
 import { useSearchParams, useNavigate } from "react-router-dom";
-// 简易消息工具（UI 库未导出 message）
-const msg = {
-  success: (s: string) => console.log(`✅ ${s}`),
-  error: (s: string) => console.error(`❌ ${s}`),
-  warning: (s: string) => console.warn(`⚠️ ${s}`),
-  info: (s: string) => console.info(`ℹ️ ${s}`),
-};
+import { message } from "antd";
+import { useQuery } from "@tanstack/react-query";
 import type {
   EditingTemplate,
   TemplateCategory,
@@ -20,26 +15,28 @@ import type {
 import {
   getEditingTemplates,
   getEditingTemplate,
+  createEditingTemplate,
   updateEditingTemplate,
   getTemplateCategories,
   generateFromTemplate,
   MODE_LABELS,
 } from "@/api/editingPlanner";
-import type {
-  EditPlanClip,
-  EditPlanGeneration,
-  MediaAsset,
-} from "@/api/editPlans";
+import type { EditPlanGeneration, MediaAsset } from "@/api/editPlans";
 import {
   getMediaAssets,
   getEditPlanGenerations,
-  MATERIAL_TYPE_LABELS,
-  aiRecommendClips,
   generateCover,
 } from "@/api/editPlans";
 import { useUndoRedo } from "./hooks/useUndoRedo";
 import type { TaskItem } from "@/api/tasks";
 import { createGenerationTask, getTask, retryTask } from "@/api/tasks";
+import type { ClipData, ClipType } from "./types";
+import {
+  ensureDefaultLibrary,
+  getAssetsByKind,
+  type AssetItem,
+} from "@/api/assets";
+import { getOrCreateDefaultProject } from "@/api/projects";
 
 import MediaPanel from "./components/MediaPanel";
 import PreviewPlayer from "./components/PreviewPlayer";
@@ -51,14 +48,6 @@ import type { GenPhase } from "./components/GenerationProgressModal";
 import GenerationHistoryModal from "./components/GenerationHistoryModal";
 import "./EditingPlanner.css";
 
-/* ──────────── 类型 ──────────── */
-
-interface ClipData extends EditPlanClip {
-  name: string;
-  thumbnail?: string;
-  assetName?: string;
-}
-
 /* ──────────── 常量 ──────────── */
 
 const MODE_LIST: { key: TemplateMode; label: string; icon: string }[] = [
@@ -106,7 +95,7 @@ const EditingPlanner: React.FC = () => {
   const [selectedClipId, setSelectedClipId] = useState<string | null>(null);
 
   /* ── AI 操作状态 ── */
-  const [aiRecommendLoading, setAiRecommendLoading] = useState(false);
+
   const [aiCoverLoading, setAiCoverLoading] = useState(false);
 
   /* ── 封面方案 ── */
@@ -119,7 +108,7 @@ const EditingPlanner: React.FC = () => {
 
   /* ── 标题/字幕/BGM 设置 ── */
   const [titleSettings, setTitleSettings] = useState({
-    aiAutoSelect: true,
+    aiAutoSelect: false,
     title: "",
     position: "top",
     font: "思源黑体",
@@ -128,6 +117,7 @@ const EditingPlanner: React.FC = () => {
     italic: false,
     stroke: true,
     shadow: true,
+    color: "#ffffff",
   });
 
   const [subtitleSettings, setSubtitleSettings] = useState({
@@ -160,6 +150,11 @@ const EditingPlanner: React.FC = () => {
 
   /* ── 素材库 ── */
   const [mediaAssets, setMediaAssets] = useState<MediaAsset[]>([]);
+  const [selectedAssetIds, setSelectedAssetIds] = useState<string[]>([]);
+
+  const handleAssetSelect = (ids: string[]) => {
+    setSelectedAssetIds(ids);
+  };
 
   /* ── 生成历史 ── */
   const [genHistoryOpen, setGenHistoryOpen] = useState(false);
@@ -169,6 +164,37 @@ const EditingPlanner: React.FC = () => {
   /* ── 播放 ── */
   const [isPlaying, setIsPlaying] = useState(false);
 
+  /* ── 配音素材（queryKey 与 VoiceMaterialLibrary 共享缓存） ── */
+  const voiceMaterialsQuery = useQuery({
+    queryKey: ["assets", "voice"],
+    queryFn: async () => {
+      const project = await getOrCreateDefaultProject();
+      await ensureDefaultLibrary({ project_id: project.id, kind: "voice" });
+      const assets = await getAssetsByKind("voice");
+      return assets;
+    },
+    staleTime: 30_000,
+  });
+  const voiceMaterials: AssetItem[] = voiceMaterialsQuery.data ?? [];
+
+  /** 为片段选择配音素材 */
+  const handleClipVoiceSelect = useCallback(
+    (clipId: string, asset: AssetItem | null) => {
+      setClips((prev) =>
+        prev.map((c) =>
+          c.id === clipId
+            ? {
+                ...c,
+                voice_asset_id: asset?.id ?? undefined,
+                voice_file_url: asset?.file_url ?? undefined,
+              }
+            : c,
+        ),
+      );
+    },
+    [setClips],
+  );
+
   /* ──────────── 加载 ──────────── */
 
   /**
@@ -187,7 +213,7 @@ const EditingPlanner: React.FC = () => {
       setCategories(cats);
       setMediaAssets(assets);
     } catch {
-      msg.error("加载模板失败");
+      message.error("加载模板失败");
     } finally {
       setLoadingTemplates(false);
     }
@@ -211,14 +237,13 @@ const EditingPlanner: React.FC = () => {
         const mapped: ClipData[] = tpl.segments.map((seg, idx) => ({
           id: seg.id || `seg-${idx}`,
           template_segment_id: seg.id || `seg-${idx}`,
-          name: `片段${idx + 1} · ${MATERIAL_TYPE_LABELS[seg.material_type || "video"] || "视频"}`,
-          material_type:
-            (seg.material_type as ClipData["material_type"]) || "video",
-          script_text: "",
+          type: (seg.material_type === "voiceover"
+            ? "voice"
+            : "pip") as ClipType,
           duration: (seg.duration_min + seg.duration_max) / 2,
+          startOffset: 0,
+          script_text: "",
           order: seg.segment_order,
-          assetName: undefined,
-          transition: { type: "none", duration: 0 },
         }));
         resetClips(mapped);
 
@@ -229,6 +254,7 @@ const EditingPlanner: React.FC = () => {
           position: tpl.title_config.position,
           font: tpl.title_config.font_preset,
           size: tpl.title_config.font_size,
+          color: tpl.title_config.font_color || "#ffffff",
         }));
         setSubtitleSettings({
           enabled: tpl.subtitle_config.enabled,
@@ -244,8 +270,8 @@ const EditingPlanner: React.FC = () => {
         setDraftCategory(tpl.category);
         setDraftTags(tpl.tags.join(", "));
       })
-      .catch(() => msg.error("加载模板详情失败"));
-  }, [loadedTemplateId]);
+      .catch(() => message.error("加载模板详情失败"));
+  }, [loadedTemplateId, resetClips]);
 
   /* ──────────── 计算 ──────────── */
 
@@ -298,91 +324,32 @@ const EditingPlanner: React.FC = () => {
     );
   };
 
-  const handleAssetDropToClip = (clipId: string, asset: MediaAsset) => {
-    handleClipUpdate(clipId, {
-      media_asset_id: asset.id,
-      assetName: asset.name,
-      thumbnail: asset.thumbnail_url,
-    });
-  };
-
-  const handleAssetDropToTimeline = (asset: MediaAsset) => {
-    const newClip: ClipData = {
-      id: `clip-${Date.now()}`,
-      template_segment_id: "",
-      name: `片段${clips.length + 1} · ${MATERIAL_TYPE_LABELS[asset.type]}`,
-      material_type:
-        asset.type === "audio"
-          ? "voiceover"
-          : asset.type === "image"
-            ? "image"
-            : "video",
-      script_text: "",
-      duration: asset.duration || 5,
-      order: clips.length,
-      media_asset_id: asset.id,
-      assetName: asset.name,
-      thumbnail: asset.thumbnail_url,
-      transition: { type: "none", duration: 0 },
-    };
-    setClips((prev) => [...prev, newClip]);
-  };
-
   /**
-   * AI 推荐片段方案
-   * 将当前已绑定素材的片段 ID 列表 + 编辑模式 + 目标时长发送给后端
-   * 返回的推荐结果直接替换当前片段列表（不保留原手动编排）
+   * 添加片段（不绑定任何素材）
+   * 片段 = 时间规划 + 类型标记
    */
-  const handleAiRecommend = async () => {
-    if (!loadedTemplateId) {
-      msg.warning("请先加载模板");
-      return;
-    }
-    const assetIds = clips
-      .map((c) => c.media_asset_id)
-      .filter((v): v is string => !!v);
-    if (assetIds.length === 0) {
-      msg.warning("请先添加素材到片段");
-      return;
-    }
-    setAiRecommendLoading(true);
-    try {
-      const res = await aiRecommendClips(loadedTemplateId, {
-        asset_ids: assetIds,
-        editing_mode: currentMode,
-        target_duration: totalDuration || 30,
-      });
-      const mapped: ClipData[] = res.clips.map((c, i) => ({
-        id: `clip-ai-${i}-${Date.now()}`,
-        template_segment_id: "",
-        name: `片段${i + 1} · ${c.clip_type}`,
-        material_type: (c.clip_type as ClipData["material_type"]) || "video",
-        script_text: c.text_content || "",
-        duration: c.duration || 5,
-        order: c.order ?? i,
-        media_asset_id: c.asset_id || undefined,
-        assetName: undefined,
-        transition: { type: "none", duration: 0 },
-      }));
-      setClips(mapped);
-      msg.success(`AI 推荐了 ${mapped.length} 个片段`);
-    } catch {
-      msg.error("AI 推荐失败");
-    } finally {
-      setAiRecommendLoading(false);
-    }
-  };
+  const handleAddClip = useCallback(
+    (type: ClipType, duration: number) => {
+      const newClip: ClipData = {
+        id: `clip-${Date.now()}`,
+        type,
+        duration,
+        startOffset: 0,
+        order: clips.length,
+      };
+      setClips((prev) => [...prev, newClip]);
+    },
+    [clips.length, setClips],
+  );
 
   /* AI 封面生成 */
   const handleAiGenerateCover = async (
     coverType: "ai_frame" | "ai_regenerate",
   ) => {
     if (!loadedTemplateId) return;
-    const assetIds = clips
-      .map((c) => c.media_asset_id)
-      .filter((v): v is string => !!v);
+    const assetIds = selectedAssetIds;
     if (assetIds.length === 0) {
-      msg.warning("请先添加素材");
+      message.warning("请先在素材库中选择素材");
       return;
     }
     setAiCoverLoading(true);
@@ -394,29 +361,24 @@ const EditingPlanner: React.FC = () => {
       setCurrentCoverScheme(
         coverType === "ai_frame" ? "ai_frame" : "ai_reselect",
       );
-      msg.success("AI 封面生成成功");
+      message.success("AI 封面生成成功");
     } catch {
-      msg.error("AI 封面生成失败");
+      message.error("AI 封面生成失败");
     } finally {
       setAiCoverLoading(false);
     }
   };
 
-  /* 保存 */
+  /* 保存 — 无论是否已加载模板，都打开保存弹窗；未加载时创建新模板 */
   const handleOpenSaveModal = () => {
-    if (!loadedTemplateId) {
-      msg.warning("请先加载一个模板");
-      return;
-    }
     setSaveModalOpen(true);
   };
 
   const handleSave = async () => {
     if (!draftName.trim()) {
-      msg.warning("请输入模板名称");
+      message.warning("请输入模板名称");
       return;
     }
-    if (!loadedTemplateId) return;
     setSaveLoading(true);
     try {
       const payload: SaveTemplatePayload = {
@@ -431,7 +393,7 @@ const EditingPlanner: React.FC = () => {
           ai_auto_select: titleSettings.aiAutoSelect,
           content: titleSettings.title,
           font_preset: titleSettings.font,
-          font_color: "#ffffff",
+          font_color: titleSettings.color,
           font_size: titleSettings.size,
           position: titleSettings.position,
         },
@@ -452,15 +414,19 @@ const EditingPlanner: React.FC = () => {
           segment_order: i,
           duration_min: Math.max(1, c.duration - 2),
           duration_max: c.duration + 2,
-          material_type: c.material_type,
+          material_type: c.type === "voice" ? "voiceover" : "video",
         })),
       };
-      await updateEditingTemplate(loadedTemplateId, payload);
-      msg.success("模板保存成功");
+      if (loadedTemplateId) {
+        await updateEditingTemplate(loadedTemplateId, payload);
+      } else {
+        await createEditingTemplate(payload);
+      }
+      message.success(loadedTemplateId ? "模板保存成功" : "模板创建成功");
       setSaveModalOpen(false);
       loadTemplates();
     } catch {
-      msg.error("保存失败");
+      message.error("保存失败");
     } finally {
       setSaveLoading(false);
     }
@@ -472,16 +438,13 @@ const EditingPlanner: React.FC = () => {
    * 一键生成页面从 params 解析配置，无需重复请求接口
    */
   const handleGoToGenerate = () => {
-    if (!loadedTemplateId) {
-      msg.warning("请先加载一个模板");
-      return;
-    }
     const planConfig = {
       title_config: {
         ai_auto_select: titleSettings.aiAutoSelect,
         content: titleSettings.title,
         position: titleSettings.position,
         font_preset: titleSettings.font,
+        font_color: titleSettings.color,
         font_size: titleSettings.size,
         bold: titleSettings.bold,
         italic: titleSettings.italic,
@@ -503,16 +466,18 @@ const EditingPlanner: React.FC = () => {
       total_duration: totalDuration,
       segments: clips.map((c, i) => ({
         order: i,
-        name: c.name,
-        material_type: c.material_type,
+        material_type: c.type === "voice" ? "voiceover" : "video",
         duration: c.duration,
-        media_asset_id: c.media_asset_id,
         template_segment_id: c.template_segment_id,
         script_text: c.script_text,
+        voice_asset_id: c.voice_asset_id,
+        voice_file_url: c.voice_file_url,
       })),
     };
     const params = new URLSearchParams();
-    params.set("edit_plan_id", loadedTemplateId);
+    if (loadedTemplateId) {
+      params.set("edit_plan_id", loadedTemplateId);
+    }
     params.set("plan_config", JSON.stringify(planConfig));
     navigate(`/app/generate?${params.toString()}`);
   };
@@ -530,21 +495,23 @@ const EditingPlanner: React.FC = () => {
       await generateFromTemplate(loadedTemplateId, {
         voiceover_duration: voiceoverDuration || totalDuration,
       });
+      // 收集所有 voice 类型片段的配音素材 ID
+      const voiceIds = clips
+        .filter((c) => c.type === "voice" && c.voice_asset_id)
+        .map((c) => c.voice_asset_id as string);
       const res = await createGenerationTask({
         template_id: loadedTemplateId,
-        asset_ids: clips
-          .map((c) => c.media_asset_id)
-          .filter((v): v is string => !!v),
+        asset_ids: [],
         title_ids: [],
-        voice_ids: [],
+        voice_ids: voiceIds,
       });
       /* 创建接口返回的是精简响应，需查询完整 TaskItem 用于轮询 */
       const task = await getTask(res.id);
       setGenTask(task);
       setGenPhase("progress");
-      msg.info("生成任务已创建");
+      message.info("生成任务已创建");
     } catch {
-      msg.error("创建生成任务失败");
+      message.error("创建生成任务失败");
     } finally {
       setGenSubmitting(false);
     }
@@ -583,7 +550,7 @@ const EditingPlanner: React.FC = () => {
       setGenTask(t);
       setGenPhase("progress");
     } catch {
-      msg.error("重试失败");
+      message.error("重试失败");
     } finally {
       setGenSubmitting(false);
     }
@@ -599,7 +566,7 @@ const EditingPlanner: React.FC = () => {
   /* 查看生成历史 */
   const handleViewGenHistory = async () => {
     if (!loadedTemplateId) {
-      msg.warning("请先加载一个模板");
+      message.warning("请先加载一个模板");
       return;
     }
     setGenHistoryOpen(true);
@@ -608,7 +575,7 @@ const EditingPlanner: React.FC = () => {
       const items = await getEditPlanGenerations(loadedTemplateId);
       setGenHistory(items);
     } catch {
-      msg.error("加载生成历史失败");
+      message.error("加载生成历史失败");
     } finally {
       setGenHistoryLoading(false);
     }
@@ -688,6 +655,9 @@ const EditingPlanner: React.FC = () => {
           onLoadTemplate={handleLoadTemplate}
           onSearchChange={setSearchQuery}
           onFilterChange={setCurrentFilter}
+          mediaAssets={mediaAssets}
+          onAssetSelect={handleAssetSelect}
+          selectedAssetIds={selectedAssetIds}
         />
 
         {/* 中栏 flex-1 */}
@@ -700,6 +670,8 @@ const EditingPlanner: React.FC = () => {
             currentCoverScheme={currentCoverScheme}
             coverSchemes={COVER_SCHEMES}
             aiCoverLoading={aiCoverLoading}
+            titleSettings={titleSettings}
+            subtitleSettings={subtitleSettings}
             onClipSelect={handleClipSelect}
             onCoverSchemeChange={setCurrentCoverScheme}
             onPlayPause={() => setIsPlaying(!isPlaying)}
@@ -714,9 +686,7 @@ const EditingPlanner: React.FC = () => {
             onClipSelect={handleClipSelect}
             onClipReorder={handleClipReorder}
             onClipRemove={handleClipRemove}
-            onAssetDropToTimeline={handleAssetDropToTimeline}
-            onAssetDropToClip={handleAssetDropToClip}
-            mediaAssets={mediaAssets}
+            onAddClip={handleAddClip}
           />
         </div>
 
@@ -729,7 +699,6 @@ const EditingPlanner: React.FC = () => {
           clipsCount={clips.length}
           totalDuration={totalDuration}
           currentMode={currentMode}
-          aiRecommendLoading={aiRecommendLoading}
           onTitleSettingsChange={(partial) =>
             setTitleSettings((prev) => ({ ...prev, ...partial }))
           }
@@ -740,7 +709,10 @@ const EditingPlanner: React.FC = () => {
             setBgmSettings((prev) => ({ ...prev, ...partial }))
           }
           onClipUpdate={handleClipUpdate}
-          onAiRecommend={handleAiRecommend}
+          voiceMaterials={voiceMaterials}
+          voiceMaterialsLoading={voiceMaterialsQuery.isLoading}
+          onRefreshVoiceMaterials={() => voiceMaterialsQuery.refetch()}
+          onClipVoiceSelect={handleClipVoiceSelect}
         />
       </div>
 

apps/web/src/pages/editing-planner/components/ClipPropertiesPanel.tsx

@@ -2,21 +2,11 @@
  * 右栏设置面板 — V8 原型 1:1 还原
  * 标题设置(AI toggle) + 字幕设置 + BGM设置 + 片段详情
  */
-import React from "react";
+import React, { useRef, useState, useCallback } from "react";
+import { useNavigate } from "react-router-dom";
 import type { TemplateMode } from "@/api/editingPlanner";
-import { MATERIAL_TYPE_LABELS } from "@/api/editPlans";
-
-interface ClipData {
-  id: string;
-  name: string;
-  duration: number;
-  material_type: "video" | "image" | "audio" | "voiceover";
-  thumbnail?: string;
-  assetName?: string;
-  media_asset_id?: string;
-  template_segment_id?: string;
-  script_text?: string;
-}
+import type { ClipData, ClipType } from "../types";
+import type { AssetItem } from "@/api/assets";
 
 interface TitleSettings {
   aiAutoSelect: boolean;
@@ -28,6 +18,7 @@ interface TitleSettings {
   italic: boolean;
   stroke: boolean;
   shadow: boolean;
+  color: string;
 }
 
 interface SubtitleSettings {
@@ -50,12 +41,18 @@ interface ClipPropertiesPanelProps {
   clipsCount: number;
   totalDuration: number;
   currentMode: TemplateMode;
-  aiRecommendLoading: boolean;
   onTitleSettingsChange: (partial: Partial<TitleSettings>) => void;
   onSubtitleSettingsChange: (partial: Partial<SubtitleSettings>) => void;
   onBgmSettingsChange: (partial: Partial<BgmSettings>) => void;
   onClipUpdate: (clipId: string, data: Partial<ClipData>) => void;
-  onAiRecommend: () => void;
+  /** 配音素材列表（从配音素材库 API 获取） */
+  voiceMaterials?: AssetItem[];
+  /** 配音素材加载中 */
+  voiceMaterialsLoading?: boolean;
+  /** 刷新配音素材列表 */
+  onRefreshVoiceMaterials?: () => void;
+  /** 为片段选择配音素材 */
+  onClipVoiceSelect?: (clipId: string, asset: AssetItem | null) => void;
 }
 
 const POSITION_OPTIONS = [
@@ -64,7 +61,15 @@ const POSITION_OPTIONS = [
   { value: "bottom", label: "底部" },
 ];
 
-const FONT_OPTIONS = ["思源黑体", "思源宋体", "苹方", "PingFang", "微软雅黑"];
+const FONT_OPTIONS = [
+  "思源黑体",
+  "思源宋体",
+  "苹方",
+  "PingFang",
+  "微软雅黑",
+  "楷体",
+  "华康俪金黑",
+];
 
 const ANIMATION_OPTIONS = [
   { value: "none", label: "无" },
@@ -81,6 +86,182 @@ const BGM_OPTIONS = [
   { value: "bgm_04", label: "🎵 科技感" },
 ];
 
+/**
+ * 标题样式预设 — 纯样式组合（颜色+描边+阴影+字重+字号）
+ * 不绑定字体，用户可自由搭配任意字体
+ * 预览统一用系统字体展示效果
+ */
+const TITLE_PRESETS = [
+  {
+    key: "classic_white",
+    label: "经典白字",
+    style: {
+      size: 28,
+      color: "#ffffff",
+      bold: true,
+      italic: false,
+      stroke: true,
+      shadow: false,
+    },
+    previewStyle: {
+      fontWeight: 700,
+      color: "#ffffff",
+      WebkitTextStroke: "1px #000000",
+      fontSize: "20px",
+    },
+  },
+  {
+    key: "black_gold",
+    label: "黑金质感",
+    style: {
+      size: 32,
+      color: "#d4a843",
+      bold: true,
+      italic: false,
+      stroke: false,
+      shadow: true,
+    },
+    previewStyle: {
+      fontWeight: 700,
+      color: "#d4a843",
+      textShadow: "1px 1px 3px rgba(0,0,0,0.8)",
+      fontSize: "20px",
+    },
+  },
+  {
+    key: "fresh_minimal",
+    label: "清新简约",
+    style: {
+      size: 24,
+      color: "#333333",
+      bold: false,
+      italic: false,
+      stroke: false,
+      shadow: false,
+    },
+    previewStyle: {
+      fontWeight: 400,
+      color: "#333333",
+      fontSize: "18px",
+    },
+  },
+  {
+    key: "variety_show",
+    label: "综艺花字",
+    style: {
+      size: 36,
+      color: "#ff4081",
+      bold: true,
+      italic: false,
+      stroke: true,
+      shadow: true,
+    },
+    previewStyle: {
+      fontWeight: 900,
+      color: "#ff4081",
+      WebkitTextStroke: "1.5px #ffffff",
+      textShadow: "2px 2px 4px rgba(0,0,0,0.5)",
+      fontSize: "22px",
+    },
+  },
+  {
+    key: "business",
+    label: "商务极简",
+    style: {
+      size: 24,
+      color: "#1a1a1a",
+      bold: false,
+      italic: false,
+      stroke: false,
+      shadow: false,
+    },
+    previewStyle: {
+      fontWeight: 400,
+      color: "#1a1a1a",
+      fontSize: "17px",
+    },
+  },
+  {
+    key: "retro_film",
+    label: "复古胶片",
+    style: {
+      size: 28,
+      color: "#e8d5b7",
+      bold: false,
+      italic: false,
+      stroke: false,
+      shadow: true,
+    },
+    previewStyle: {
+      fontWeight: 400,
+      color: "#e8d5b7",
+      textShadow: "2px 2px 6px rgba(0,0,0,0.7)",
+      fontSize: "18px",
+    },
+  },
+  {
+    key: "neon_glow",
+    label: "霓虹发光",
+    style: {
+      size: 32,
+      color: "#00e5ff",
+      bold: true,
+      italic: false,
+      stroke: false,
+      shadow: true,
+    },
+    previewStyle: {
+      fontWeight: 700,
+      color: "#00e5ff",
+      textShadow:
+        "0 0 4px #00e5ff, 0 0 8px #00e5ff, 0 0 16px rgba(0,229,255,0.5)",
+      fontSize: "20px",
+    },
+  },
+  {
+    key: "handwriting",
+    label: "手写字",
+    style: {
+      size: 28,
+      color: "#333333",
+      bold: false,
+      italic: false,
+      stroke: false,
+      shadow: true,
+    },
+    previewStyle: {
+      fontWeight: 400,
+      color: "#333333",
+      textShadow: "1px 1px 2px rgba(0,0,0,0.3)",
+      fontSize: "20px",
+    },
+  },
+];
+
+/** 判断当前设置匹配哪个预设（比较 size + color + bold/italic/stroke/shadow，不比较字体） */
+function getActivePreset(settings: TitleSettings): string | null {
+  for (const p of TITLE_PRESETS) {
+    if (
+      settings.size === p.style.size &&
+      settings.color === p.style.color &&
+      settings.bold === p.style.bold &&
+      settings.italic === p.style.italic &&
+      settings.stroke === p.style.stroke &&
+      settings.shadow === p.style.shadow
+    ) {
+      return p.key;
+    }
+  }
+  return null;
+}
+
+/** 片段类型图标/标签 */
+const CLIP_TYPE_ICONS: Record<ClipType, string> = { voice: "🎙️", pip: "🖼️" };
+const CLIP_TYPE_LABELS: Record<ClipType, string> = {
+  voice: "口播",
+  pip: "画中画",
+};
+
 const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
   selectedClip,
   titleSettings,
@@ -89,13 +270,50 @@ const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
   clipsCount,
   totalDuration,
   currentMode,
-  aiRecommendLoading,
   onTitleSettingsChange,
   onSubtitleSettingsChange,
   onBgmSettingsChange,
   onClipUpdate,
-  onAiRecommend,
+  voiceMaterials = [],
+  voiceMaterialsLoading = false,
+  onRefreshVoiceMaterials,
+  onClipVoiceSelect,
 }) => {
+  const navigate = useNavigate();
+
+  /* ── 配音试听 ── */
+  const audioRef = useRef<HTMLAudioElement | null>(null);
+  const [previewingId, setPreviewingId] = useState<string | null>(null);
+
+  /** 试听配音素材 */
+  const handlePreviewVoice = useCallback(
+    (asset: AssetItem) => {
+      // 点击同一个 → 暂停
+      if (previewingId === asset.id) {
+        audioRef.current?.pause();
+        setPreviewingId(null);
+        return;
+      }
+      // 停止上一个
+      audioRef.current?.pause();
+      const url = asset.file_url || (asset.metadata?.preview_url as string);
+      if (!url) return;
+      const audio = new Audio(url);
+      audioRef.current = audio;
+      audio.play().catch(() => {});
+      audio.onended = () => setPreviewingId(null);
+      setPreviewingId(asset.id);
+    },
+    [previewingId],
+  );
+
+  /** 从 metadata 取性别标签 */
+  const getGenderLabel = (m: AssetItem): string => {
+    const g = (m.metadata?.gender as string) || "";
+    if (g === "male") return "男";
+    if (g === "female") return "女";
+    return "";
+  };
   return (
     <div className="ep-right-panel">
       {/* ═══ 标题设置 ═══ */}
@@ -172,6 +390,42 @@ const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
               </div>
             </div>
 
+            <div className="ep-field">
+              <label className="ep-field-label">预设样式</label>
+              <div className="ep-title-presets-grid">
+                {TITLE_PRESETS.map((p) => {
+                  const isActive = getActivePreset(titleSettings) === p.key;
+                  return (
+                    <button
+                      key={p.key}
+                      className={`ep-title-preset-card${isActive ? " active" : ""}`}
+                      onClick={() =>
+                        onTitleSettingsChange({
+                          size: p.style.size,
+                          color: p.style.color,
+                          bold: p.style.bold,
+                          italic: p.style.italic,
+                          stroke: p.style.stroke,
+                          shadow: p.style.shadow,
+                        })
+                      }
+                      title={p.label}
+                    >
+                      <span
+                        className="ep-title-preset-preview-text"
+                        style={p.previewStyle}
+                      >
+                        标题
+                      </span>
+                      <span className="ep-title-preset-card-label">
+                        {p.label}
+                      </span>
+                    </button>
+                  );
+                })}
+              </div>
+            </div>
+
             <div className="ep-field">
               <label className="ep-field-label">样式</label>
               <div className="ep-style-btns">
@@ -345,15 +599,36 @@ const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
           </div>
 
           <div className="ep-clip-detail">
-            <div className="ep-clip-detail-header">
-              <span className="ep-clip-detail-name">{selectedClip.name}</span>
-              <span className="ep-clip-detail-type">
-                {MATERIAL_TYPE_LABELS[selectedClip.material_type] || "视频"}
-              </span>
+            {/* 类型选择器 */}
+            <div className="ep-clip-detail-field">
+              <div className="ep-clip-detail-label">类型</div>
+              <div className="ep-clip-type-selector">
+                {(["voice", "pip"] as ClipType[]).map((t) => {
+                  const disabled =
+                    currentMode === "pip"
+                      ? t !== "pip"
+                      : currentMode === "voice_over"
+                        ? t !== "voice"
+                        : false; // voice_pip 可切换
+                  return (
+                    <button
+                      key={t}
+                      className={`ep-clip-type-btn${selectedClip.type === t ? " active" : ""}${disabled ? " disabled" : ""}`}
+                      disabled={disabled}
+                      onClick={() =>
+                        !disabled && onClipUpdate(selectedClip.id, { type: t })
+                      }
+                    >
+                      {CLIP_TYPE_ICONS[t]} {CLIP_TYPE_LABELS[t]}
+                    </button>
+                  );
+                })}
+              </div>
             </div>
 
+            {/* 时长 */}
             <div className="ep-clip-detail-field">
-              <div className="ep-clip-detail-label">时长范围</div>
+              <div className="ep-clip-detail-label">时长</div>
               <div className="ep-clip-detail-row">
                 <input
                   className="ep-duration-input"
@@ -363,7 +638,10 @@ const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
                   value={selectedClip.duration}
                   onChange={(e) =>
                     onClipUpdate(selectedClip.id, {
-                      duration: Number(e.target.value),
+                      duration: Math.max(
+                        1,
+                        Math.min(120, Number(e.target.value) || 1),
+                      ),
                     })
                   }
                 />
@@ -371,31 +649,127 @@ const ClipPropertiesPanel: React.FC<ClipPropertiesPanelProps> = ({
               </div>
             </div>
 
-            {selectedClip.assetName && (
+            {/* 素材起始时间 — 仅 voice 类型显示 */}
+            {selectedClip.type === "voice" && (
               <div className="ep-clip-detail-field">
-                <div className="ep-clip-detail-label">关联素材</div>
-                <div className="ep-clip-detail-value">
-                  <span className="ep-clip-detail-linked">
-                    {selectedClip.assetName}
-                  </span>
+                <div className="ep-clip-detail-label">素材起始时间</div>
+                <div className="ep-clip-detail-row">
+                  <input
+                    className="ep-duration-input"
+                    type="number"
+                    min={0}
+                    max={9999}
+                    step={0.1}
+                    value={selectedClip.startOffset}
+                    onChange={(e) =>
+                      onClipUpdate(selectedClip.id, {
+                        startOffset: Math.max(
+                          0,
+                          Math.min(9999, Number(e.target.value) || 0),
+                        ),
+                      })
+                    }
+                  />
+                  <span className="ep-clip-detail-unit">秒</span>
                 </div>
               </div>
             )}
 
-            <div className="ep-clip-detail-field">
-              <div className="ep-clip-detail-label">素材类型</div>
-              <div className="ep-clip-detail-value">
-                {MATERIAL_TYPE_LABELS[selectedClip.material_type] || "视频"}
-              </div>
-            </div>
+            {/* 配音素材选择 — 仅 voice 类型显示 */}
+            {selectedClip.type === "voice" && (
+              <div className="ep-clip-detail-field">
+                <div className="ep-clip-detail-label">
+                  配音素材
+                  {onRefreshVoiceMaterials && (
+                    <button
+                      type="button"
+                      className="ep-voice-refresh-btn"
+                      title="刷新配音列表"
+                      onClick={() => onRefreshVoiceMaterials()}
+                      disabled={voiceMaterialsLoading}
+                    >
+                      {voiceMaterialsLoading ? "⏳" : "🔄"}
+                    </button>
+                  )}
+                </div>
 
-            <button
-              className="ep-ai-recommend-btn"
-              onClick={onAiRecommend}
-              disabled={aiRecommendLoading}
-            >
-              {aiRecommendLoading ? "⏳ AI 分析中..." : "✨ AI 推荐素材"}
-            </button>
+                {voiceMaterialsLoading && voiceMaterials.length === 0 ? (
+                  <div className="ep-voice-loading">加载中...</div>
+                ) : (
+                  <>
+                    <div className="ep-voice-select-row">
+                      <select
+                        className="ep-clip-detail-select"
+                        value={selectedClip.voice_asset_id ?? ""}
+                        onChange={(e) => {
+                          const assetId = e.target.value;
+                          if (!onClipVoiceSelect) return;
+                          if (!assetId) {
+                            onClipVoiceSelect(selectedClip.id, null);
+                          } else {
+                            const asset = voiceMaterials.find(
+                              (m) => m.id === assetId,
+                            );
+                            if (asset)
+                              onClipVoiceSelect(selectedClip.id, asset);
+                          }
+                          // 切换选择时停止试听
+                          audioRef.current?.pause();
+                          setPreviewingId(null);
+                        }}
+                      >
+                        <option value="">未选择</option>
+                        {voiceMaterials.map((m) => {
+                          const gender = getGenderLabel(m);
+                          const label = gender
+                            ? `${m.name}（${gender}）`
+                            : m.name;
+                          return (
+                            <option key={m.id} value={m.id}>
+                              {label}
+                            </option>
+                          );
+                        })}
+                      </select>
+                      {/* 试听按钮 */}
+                      {selectedClip.voice_asset_id && (
+                        <button
+                          type="button"
+                          className="ep-voice-preview-btn"
+                          title={
+                            previewingId === selectedClip.voice_asset_id
+                              ? "暂停"
+                              : "试听"
+                          }
+                          onClick={() => {
+                            const asset = voiceMaterials.find(
+                              (m) => m.id === selectedClip.voice_asset_id,
+                            );
+                            if (asset) handlePreviewVoice(asset);
+                          }}
+                        >
+                          {previewingId === selectedClip.voice_asset_id
+                            ? "⏸"
+                            : "▶️"}
+                        </button>
+                      )}
+                    </div>
+                    {voiceMaterials.length === 0 && (
+                      <div className="ep-voice-empty">
+                        暂无配音素材，请先上传
+                      </div>
+                    )}
+                  </>
+                )}
+
+                <button
+                  className="ep-voice-upload-btn"
+                  onClick={() => navigate("/app/voice-materials")}
+                >
+                  + 上传新配音
+                </button>
+              </div>
+            )}
           </div>
         </div>
       )}

apps/web/src/pages/editing-planner/components/MediaPanel.tsx

@@ -1,10 +1,12 @@
 /**
- * 左侧模板面板 — V8 原型 1:1 还原
- * 纯模板列表 + chip 分类筛选（无素材 Tab）
+ * 左侧面板 — V8 原型 1:1 还原
+ * Tab 切换：模板列表 + 素材库
  */
-import React from "react";
+import React, { useState } from "react";
 import type { EditingTemplate } from "@/api/editingPlanner";
 import { MODE_LABELS } from "@/api/editingPlanner";
+import type { MediaAsset } from "@/api/editPlans";
+import AssetSelector from "@/components/AssetSelector/AssetSelector";
 
 interface MediaPanelProps {
   templates: EditingTemplate[];
@@ -16,6 +18,10 @@ interface MediaPanelProps {
   onLoadTemplate: (id: string) => void;
   onSearchChange: (q: string) => void;
   onFilterChange: (f: string) => void;
+  // 素材相关
+  mediaAssets?: MediaAsset[];
+  onAssetSelect?: (ids: string[]) => void;
+  selectedAssetIds?: string[];
 }
 
 const MediaPanel: React.FC<MediaPanelProps> = ({
@@ -28,80 +34,117 @@ const MediaPanel: React.FC<MediaPanelProps> = ({
   onLoadTemplate,
   onSearchChange,
   onFilterChange,
+  mediaAssets = [],
+  onAssetSelect,
+  selectedAssetIds = [],
 }) => {
+  const [activeTab, setActiveTab] = useState<"templates" | "assets">(
+    "templates",
+  );
+
   return (
     <div className="ep-left-panel">
-      {/* 标题栏 */}
-      <div className="ep-left-header">
-        <span className="ep-left-title">📋 已保存模板</span>
+      {/* Tab 切换 */}
+      <div className="ep-left-tabs">
+        <button
+          className={`ep-left-tab ${activeTab === "templates" ? "active" : ""}`}
+          onClick={() => setActiveTab("templates")}
+        >
+          📋 模板
+        </button>
+        <button
+          className={`ep-left-tab ${activeTab === "assets" ? "active" : ""}`}
+          onClick={() => setActiveTab("assets")}
+        >
+          📁 素材
+        </button>
       </div>
 
-      {/* 搜索 */}
-      <div className="ep-search-wrap ep-media-panel-inner">
-        <span className="ep-search-icon">🔍</span>
-        <input
-          className="ep-search-input"
-          placeholder="搜索模板..."
-          value={searchQuery}
-          onChange={(e) => onSearchChange(e.target.value)}
-        />
-      </div>
-
-      {/* Chip 分类筛选 */}
-      <div className="ep-filter-chips">
-        {filterCategories.map((cat) => (
-          <button
-            key={cat}
-            className={`ep-filter-chip ${currentFilter === cat ? "active" : ""}`}
-            onClick={() => onFilterChange(cat)}
-          >
-            {cat}
-          </button>
-        ))}
-      </div>
-
-      {/* 模板列表 */}
-      <div className="ep-template-list">
-        {loading ? (
-          <div className="ep-loading">
-            <span>⏳</span>
-            <span>加载中...</span>
+      {/* 模板 Tab */}
+      {activeTab === "templates" && (
+        <>
+          {/* 搜索 */}
+          <div className="ep-search-wrap ep-media-panel-inner">
+            <span className="ep-search-icon">🔍</span>
+            <input
+              className="ep-search-input"
+              placeholder="搜索模板..."
+              value={searchQuery}
+              onChange={(e) => onSearchChange(e.target.value)}
+            />
           </div>
-        ) : templates.length === 0 ? (
-          <div className="ep-empty">
-            <span>📭</span>
-            <span>暂无模板</span>
+
+          {/* Chip 分类筛选 */}
+          <div className="ep-filter-chips">
+            {filterCategories.map((cat) => (
+              <button
+                key={cat}
+                className={`ep-filter-chip ${currentFilter === cat ? "active" : ""}`}
+                onClick={() => onFilterChange(cat)}
+              >
+                {cat}
+              </button>
+            ))}
           </div>
-        ) : (
-          templates.map((tpl) => (
-            <div
-              key={tpl.id}
-              className={`ep-template-card ${loadedTemplateId === tpl.id ? "active" : ""}`}
-              onClick={() => onLoadTemplate(tpl.id)}
-            >
-              <div className="ep-template-card-header">
-                <span className="ep-template-card-name">{tpl.name}</span>
-                <span className="ep-template-card-mode">
-                  {MODE_LABELS[tpl.mode]}
-                </span>
+
+          {/* 模板列表 */}
+          <div className="ep-template-list">
+            {loading ? (
+              <div className="ep-loading">
+                <span>⏳</span>
+                <span>加载中...</span>
               </div>
-              <div className="ep-template-card-meta">
-                <span>⏱️ {tpl.estimated_duration}s</span>
-                <span>📐 {tpl.segments.length}片段</span>
+            ) : templates.length === 0 ? (
+              <div className="ep-empty">
+                <span>📭</span>
+                <span>暂无模板</span>
               </div>
-              {tpl.tags.length > 0 && (
-                <div className="ep-template-card-tags">
-                  {tpl.tags.map((tag) => (
-                    <span key={tag} className="ep-template-tag">
-                      {tag}
+            ) : (
+              templates.map((tpl) => (
+                <div
+                  key={tpl.id}
+                  className={`ep-template-card ${loadedTemplateId === tpl.id ? "active" : ""}`}
+                  onClick={() => onLoadTemplate(tpl.id)}
+                >
+                  <div className="ep-template-card-header">
+                    <span className="ep-template-card-name">{tpl.name}</span>
+                    <span className="ep-template-card-mode">
+                      {MODE_LABELS[tpl.mode]}
                     </span>
-                  ))}
+                  </div>
+                  <div className="ep-template-card-meta">
+                    <span>⏱️ {tpl.estimated_duration}s</span>
+                    <span>📐 {tpl.segments.length}片段</span>
+                  </div>
+                  {tpl.tags.length > 0 && (
+                    <div className="ep-template-card-tags">
+                      {tpl.tags.map((tag) => (
+                        <span key={tag} className="ep-template-tag">
+                          {tag}
+                        </span>
+                      ))}
+                    </div>
+                  )}
                 </div>
-              )}
-            </div>
-          ))
-        )}
-      </div>
+              ))
+            )}
+          </div>
+        </>
+      )}
+
+      {/* 素材 Tab */}
+      {activeTab === "assets" && (
+        <div className="ep-assets-tab">
+          <AssetSelector
+            assets={mediaAssets}
+            selectedIds={selectedAssetIds}
+            onSelectionChange={onAssetSelect}
+            showQualityFilter={false}
+            showBatchSelect={false}
+            compact
+          />
+        </div>
+      )}
     </div>
   );
 };

apps/web/src/pages/editing-planner/components/PreviewPlayer.tsx

@@ -4,21 +4,33 @@
  * 封面右侧竖排4个方案按钮
  */
 import React from "react";
-
-interface ClipData {
-  id: string;
-  name: string;
-  duration: number;
-  material_type: string;
-  thumbnail?: string;
-  assetName?: string;
-}
+import type { ClipData, ClipType } from "../types";
 
 interface CoverScheme {
   key: string;
   label: string;
 }
 
+interface TitleSettings {
+  aiAutoSelect: boolean;
+  title: string;
+  position: string;
+  font: string;
+  size: number;
+  bold: boolean;
+  italic: boolean;
+  stroke: boolean;
+  shadow: boolean;
+}
+
+interface SubtitleSettings {
+  enabled: boolean;
+  position: string;
+  font: string;
+  size: number;
+  animation: string;
+}
+
 interface PreviewPlayerProps {
   clips: ClipData[];
   selectedClipId: string | null;
@@ -26,17 +38,22 @@ interface PreviewPlayerProps {
   currentCoverScheme: string;
   coverSchemes: CoverScheme[];
   aiCoverLoading: boolean;
+  titleSettings?: TitleSettings;
+  subtitleSettings?: SubtitleSettings;
   onClipSelect: (clipId: string) => void;
   onCoverSchemeChange: (scheme: string) => void;
   onPlayPause: () => void;
   onAiGenerateCover: (coverType: "ai_frame" | "ai_regenerate") => void;
 }
 
-const MATERIAL_ICONS: Record<string, string> = {
-  video: "🎬",
-  image: "🖼️",
-  audio: "🎵",
-  voiceover: "🎙️",
+const CLIP_TYPE_ICONS: Record<ClipType, string> = {
+  voice: "🎙️",
+  pip: "🖼️",
+};
+
+const CLIP_TYPE_LABELS: Record<ClipType, string> = {
+  voice: "口播",
+  pip: "画中画",
 };
 
 const PreviewPlayer: React.FC<PreviewPlayerProps> = ({
@@ -46,6 +63,8 @@ const PreviewPlayer: React.FC<PreviewPlayerProps> = ({
   currentCoverScheme,
   coverSchemes,
   aiCoverLoading,
+  titleSettings,
+  subtitleSettings,
   onCoverSchemeChange,
   onPlayPause,
   onAiGenerateCover,
@@ -73,11 +92,72 @@ const PreviewPlayer: React.FC<PreviewPlayerProps> = ({
                   style={{ width: isPlaying ? "45%" : "0%" }}
                 />
               </div>
-              <div className="ep-phone-clip-label">{displayClip.name}</div>
+              <div className="ep-phone-clip-label">
+                {CLIP_TYPE_ICONS[displayClip.type] || "🎬"}{" "}
+                {CLIP_TYPE_LABELS[displayClip.type] || "片段"}
+              </div>
             </>
           ) : (
             <span className="ep-phone-empty-hint">暂无片段</span>
           )}
+
+          {/* 标题实时预览 */}
+          {titleSettings &&
+            !titleSettings.aiAutoSelect &&
+            titleSettings.title && (
+              <div
+                className="ep-preview-title"
+                style={{
+                  fontSize: `${Math.min(titleSettings.size, 20)}px`,
+                  fontFamily: titleSettings.font,
+                  fontWeight: titleSettings.bold ? "bold" : "normal",
+                  fontStyle: titleSettings.italic ? "italic" : "normal",
+                  textShadow: titleSettings.shadow
+                    ? "2px 2px 4px rgba(0,0,0,0.5)"
+                    : "none",
+                  WebkitTextStroke: titleSettings.stroke
+                    ? "1px rgba(0,0,0,0.6)"
+                    : "none",
+                  top:
+                    titleSettings.position === "top"
+                      ? "8px"
+                      : titleSettings.position === "center"
+                        ? "50%"
+                        : "auto",
+                  bottom: titleSettings.position === "bottom" ? "30px" : "auto",
+                  transform:
+                    titleSettings.position === "center"
+                      ? "translateY(-50%)"
+                      : "none",
+                }}
+              >
+                {titleSettings.title}
+              </div>
+            )}
+
+          {/* 字幕实时预览 */}
+          {subtitleSettings?.enabled && (
+            <div
+              className="ep-preview-subtitle"
+              style={{
+                fontSize: `${Math.min(subtitleSettings.size, 14)}px`,
+                fontFamily: subtitleSettings.font,
+                top:
+                  subtitleSettings.position === "top"
+                    ? "8px"
+                    : subtitleSettings.position === "center"
+                      ? "50%"
+                      : "auto",
+                bottom: subtitleSettings.position === "bottom" ? "8px" : "auto",
+                transform:
+                  subtitleSettings.position === "center"
+                    ? "translateY(-50%)"
+                    : "none",
+              }}
+            >
+              字幕预览文字
+            </div>
+          )}
         </div>
       </div>
 
@@ -86,7 +166,7 @@ const PreviewPlayer: React.FC<PreviewPlayerProps> = ({
         <div className="ep-cover-image">
           {displayClip ? (
             <span className="ep-cover-icon">
-              {MATERIAL_ICONS[displayClip.material_type] || "🎬"}
+              {CLIP_TYPE_ICONS[displayClip.type] || "🎬"}
             </span>
           ) : (
             <span>暂无封面</span>

apps/web/src/pages/editing-planner/components/TimelinePanel.tsx

@@ -1,22 +1,16 @@
 /**
- * 水平轨道时间线 — V8 原型 1:1 还原
+ * 水平轨道时间线 — 片段 = 时间规划 + 类型标记，不绑定素材
  * 时间标尺(20px) + 水平片段卡片轨道(100x100) + HTML5拖拽排序
+ * "+" 卡片 → 类型+时长选择器
  */
-import React, { useState, useRef } from "react";
-import type { MediaAsset } from "@/api/editPlans";
-
-interface ClipData {
-  id: string;
-  name: string;
-  duration: number;
-  material_type: string;
-  thumbnail?: string;
-  assetName?: string;
-  media_asset_id?: string;
-  template_segment_id?: string;
-  script_text?: string;
-  order?: number;
-}
+import React, {
+  useState,
+  useRef,
+  useCallback,
+  useEffect,
+  useLayoutEffect,
+} from "react";
+import type { ClipData, ClipType } from "../types";
 
 interface TimelinePanelProps {
   clips: ClipData[];
@@ -25,16 +19,19 @@ interface TimelinePanelProps {
   onClipSelect: (clipId: string) => void;
   onClipReorder: (fromIdx: number, toIdx: number) => void;
   onClipRemove: (clipId: string) => void;
-  onAssetDropToTimeline: (asset: MediaAsset) => void;
-  onAssetDropToClip: (clipId: string, asset: MediaAsset) => void;
-  mediaAssets: MediaAsset[];
+  onAddClip: (type: ClipType, duration: number) => void;
 }
 
-const MATERIAL_ICONS: Record<string, string> = {
-  video: "🎬",
-  image: "🖼️",
-  audio: "🎵",
-  voiceover: "🎙️",
+/** 片段类型图标 */
+const CLIP_TYPE_ICONS: Record<ClipType, string> = {
+  voice: "🎙️",
+  pip: "🖼️",
+};
+
+/** 片段类型标签 */
+const CLIP_TYPE_LABELS: Record<ClipType, string> = {
+  voice: "口播",
+  pip: "画中画",
 };
 
 const TimelinePanel: React.FC<TimelinePanelProps> = ({
@@ -44,12 +41,112 @@ const TimelinePanel: React.FC<TimelinePanelProps> = ({
   onClipSelect,
   onClipReorder,
   onClipRemove,
-  onAssetDropToTimeline,
-  onAssetDropToClip,
+  onAddClip,
 }) => {
   const [dragIdx, setDragIdx] = useState<number | null>(null);
   const [dragOverIdx, setDragOverIdx] = useState<number | null>(null);
   const dragRef = useRef<number | null>(null);
+  const [showAddPicker, setShowAddPicker] = useState(false);
+  const pickerRef = useRef<HTMLDivElement>(null);
+  const addCardRef = useRef<HTMLDivElement>(null);
+  const [pickerPos, setPickerPos] = useState<{ top: number; right: number }>({
+    top: 0,
+    right: 0,
+  });
+
+  /* ── "+" 卡片：类型+时长选择状态 ── */
+  const [addType, setAddType] = useState<ClipType>("voice");
+  const [addDuration, setAddDuration] = useState<number>(5);
+
+  /* ── 根据模式决定可选类型 ── */
+  const availableTypes: ClipType[] =
+    currentMode === "voice_over"
+      ? ["voice"]
+      : currentMode === "pip"
+        ? ["pip"]
+        : ["voice", "pip"]; // voice_pip 或默认
+
+  /* ── 面板尺寸（宽度固定，高度由 useLayoutEffect 实测） ── */
+  const PICKER_W = 240; // 面板宽度（与 CSS 一致）
+  const GAP = 6; // 面板与"+"卡片的间距
+
+  /* ── 计算 picker 初始位置（默认从"+"按钮上方弹出） ── */
+  const updatePickerPosition = useCallback(() => {
+    if (!addCardRef.current) return;
+    const rect = addCardRef.current.getBoundingClientRect();
+    const vw = window.innerWidth;
+
+    /* 垂直方向：默认向上弹出（上方空间永远比下方大） */
+    const roughHeight = 180; // 粗略估算，useLayoutEffect 会用实际高度校正
+    let top = rect.top - GAP - roughHeight;
+    if (top < 8) top = 8;
+
+    /* 水平方向：右对齐"+"卡片；太靠右超出视口则左移 */
+    let right = vw - rect.right;
+    if (rect.right - PICKER_W < 8) {
+      right = vw - PICKER_W - 8;
+    }
+
+    setPickerPos({ top, right });
+  }, []);
+
+  const handleTogglePicker = () => {
+    if (!showAddPicker) {
+      updatePickerPosition();
+    }
+    setShowAddPicker((v) => !v);
+  };
+
+  /* ── 渲染后用实际 offsetHeight 做精确边界校正（useLayoutEffect 确保 paint 前完成） ── */
+  useLayoutEffect(() => {
+    if (!showAddPicker || !pickerRef.current || !addCardRef.current) return;
+    const pickerEl = pickerRef.current;
+    const addRect = addCardRef.current.getBoundingClientRect();
+    const pickerH = pickerEl.offsetHeight; // 实际高度，不用硬编码
+    const vh = window.innerHeight;
+    const vw = window.innerWidth;
+
+    /* 默认：面板在"+"按钮上方 */
+    let top = addRect.top - GAP - pickerH;
+
+    /* 上方空间也不够（极端情况）→ 翻转到下方 */
+    if (top < 8) {
+      top = addRect.bottom + GAP;
+      /* 下方也溢出 → clamp */
+      if (top + pickerH > vh - 8) {
+        top = vh - 8 - pickerH;
+        if (top < 8) top = 8;
+      }
+    }
+
+    /* 水平方向：右对齐"+"卡片；左侧溢出保护 */
+    let right = vw - addRect.right;
+    const pickerRect = pickerEl.getBoundingClientRect();
+    if (pickerRect.left < 8) {
+      right = vw - PICKER_W - 8;
+    }
+
+    setPickerPos({ top, right });
+  }, [showAddPicker]);
+
+  /* ── 点击外部关闭添加面板 ── */
+  useEffect(() => {
+    const handleClickOutside = (e: MouseEvent) => {
+      if (pickerRef.current && !pickerRef.current.contains(e.target as Node)) {
+        setShowAddPicker(false);
+      }
+    };
+    if (showAddPicker) {
+      document.addEventListener("mousedown", handleClickOutside);
+    }
+    return () => document.removeEventListener("mousedown", handleClickOutside);
+  }, [showAddPicker]);
+
+  /* ── 确认添加片段 ── */
+  const handleConfirmAdd = () => {
+    onAddClip(addType, addDuration);
+    setShowAddPicker(false);
+  };
 
   /* ── 片段拖拽排序 ── */
   const handleDragStart = (e: React.DragEvent, idx: number) => {
@@ -80,37 +177,12 @@ const TimelinePanel: React.FC<TimelinePanelProps> = ({
       if (fromIdx !== toIdx) {
         onClipReorder(fromIdx, toIdx);
       }
-      return;
-    }
-    // 素材拖到片段上
-    const assetJson = e.dataTransfer.getData("application/x-media-asset");
-    if (assetJson) {
-      try {
-        const asset: MediaAsset = JSON.parse(assetJson);
-        onAssetDropToClip(clips[toIdx].id, asset);
-      } catch {
-        /* ignore */
-      }
-    }
-  };
-
-  /* ── 空轨道区域拖入 ── */
-  const handleEmptyDrop = (e: React.DragEvent) => {
-    e.preventDefault();
-    const assetJson = e.dataTransfer.getData("application/x-media-asset");
-    if (assetJson) {
-      try {
-        const asset: MediaAsset = JSON.parse(assetJson);
-        onAssetDropToTimeline(asset);
-      } catch {
-        /* ignore */
-      }
     }
   };
 
+  /* ── 空轨道区域不接受素材拖入 ── */
   const handleEmptyDragOver = (e: React.DragEvent) => {
     e.preventDefault();
-    e.dataTransfer.dropEffect = "copy";
   };
 
   /* ── 时间标尺 ── */
@@ -158,79 +230,149 @@ const TimelinePanel: React.FC<TimelinePanelProps> = ({
 
       {/* 一镜到底提示 */}
       {currentMode === "one_take" && (
-        <div className="ep-one-take-hint">
-          🎥 一镜到底模式：所有片段将无缝衔接，不可添加转场
-        </div>
+        <div className="ep-one-take-hint">🎥 一镜到底模式无片段</div>
       )}
 
       {/* 时间标尺 */}
-      <div className="ep-time-ruler">
-        <div
-          className="ep-time-ruler-inner"
-          style={{ width: Math.max(clips.length * 108, 300) }}
-        >
-          {rulerMarks.map((t) => (
-            <span
-              key={t}
-              className="ep-time-mark"
-              style={{
-                left:
-                  totalDuration > 0
-                    ? `${(t / totalDuration) * clips.length * 108}px`
-                    : `${t * 20}px`,
-              }}
-            >
-              {t}s
-            </span>
-          ))}
-        </div>
-      </div>
-
-      {/* 水平片段轨道 */}
-      <div
-        className="ep-clip-track"
-        onDrop={handleEmptyDrop}
-        onDragOver={handleEmptyDragOver}
-      >
-        {clips.length === 0 ? (
-          <div className="ep-track-empty">🎬 拖入素材或从模板加载片段</div>
-        ) : (
-          clips.map((clip, idx) => (
-            <div
-              key={clip.id}
-              className={`ep-clip-card ${selectedClipId === clip.id ? "selected" : ""} ${dragIdx === idx ? "dragging" : ""} ${dragOverIdx === idx ? "drag-over" : ""}`}
-              draggable
-              onDragStart={(e) => handleDragStart(e, idx)}
-              onDragOver={(e) => handleDragOver(e, idx)}
-              onDragEnd={handleDragEnd}
-              onDrop={(e) => handleDrop(e, idx)}
-              onClick={() => onClipSelect(clip.id)}
-            >
-              {/* 缩略图区域 */}
-              <div className="ep-clip-thumbnail">
-                {MATERIAL_ICONS[clip.material_type] || "🎬"}
-              </div>
-
-              {/* 片段信息 */}
-              <div className="ep-clip-info">
-                <span className="ep-clip-name">{clip.name}</span>
-                <span className="ep-clip-duration">{clip.duration}s</span>
-              </div>
-
-              {/* 删除按钮 */}
-              <button
-                className="ep-clip-remove"
-                onClick={(e) => {
-                  e.stopPropagation();
-                  onClipRemove(clip.id);
+      {currentMode !== "one_take" && (
+        <div className="ep-time-ruler">
+          <div
+            className="ep-time-ruler-inner"
+            style={{ width: Math.max(clips.length * 108, 300) }}
+          >
+            {rulerMarks.map((t) => (
+              <span
+                key={t}
+                className="ep-time-mark"
+                style={{
+                  left:
+                    totalDuration > 0
+                      ? `${(t / totalDuration) * clips.length * 108}px`
+                      : `${t * 20}px`,
                 }}
               >
-                ✕
-              </button>
+                {t}s
+              </span>
+            ))}
+          </div>
+        </div>
+      )}
+
+      {/* 水平片段轨道 */}
+      {currentMode !== "one_take" && (
+        <div className="ep-clip-track" onDragOver={handleEmptyDragOver}>
+          {clips.length === 0 ? (
+            <div className="ep-track-empty">
+              <div className="ep-track-empty-icon">🎬</div>
+              <div className="ep-track-empty-text">点击右侧 + 添加片段</div>
             </div>
-          ))
-        )}
-      </div>
+          ) : (
+            clips.map((clip, idx) => (
+              <div
+                key={clip.id}
+                className={`ep-clip-card ${selectedClipId === clip.id ? "selected" : ""} ${dragIdx === idx ? "dragging" : ""} ${dragOverIdx === idx ? "drag-over" : ""}`}
+                draggable
+                onDragStart={(e) => handleDragStart(e, idx)}
+                onDragOver={(e) => handleDragOver(e, idx)}
+                onDragEnd={handleDragEnd}
+                onDrop={(e) => handleDrop(e, idx)}
+                onClick={() => onClipSelect(clip.id)}
+              >
+                {/* 类型图标 */}
+                <div className="ep-clip-thumbnail">
+                  {CLIP_TYPE_ICONS[clip.type] || "🎬"}
+                </div>
+
+                {/* 片段信息 */}
+                <div className="ep-clip-info">
+                  <span className="ep-clip-name">
+                    {CLIP_TYPE_LABELS[clip.type] || "片段"} {idx + 1}
+                  </span>
+                  <span className="ep-clip-duration">{clip.duration}s</span>
+                </div>
+
+                {/* 删除按钮 */}
+                <button
+                  className="ep-clip-remove"
+                  onClick={(e) => {
+                    e.stopPropagation();
+                    onClipRemove(clip.id);
+                  }}
+                >
+                  ✕
+                </button>
+              </div>
+            ))
+          )}
+
+          {/* ── 轨道末尾 "+" 添加卡片 → 类型+时长选择器 ── */}
+          <div className="ep-track-add-card-wrapper">
+            <div
+              ref={addCardRef}
+              className="ep-track-add-card"
+              onClick={handleTogglePicker}
+              title="添加片段"
+            >
+              +
+            </div>
+          </div>
+        </div>
+      )}
+
+      {/* 类型+时长选择面板 — fixed 定位，不受任何父容器 overflow 裁剪 */}
+      {showAddPicker && (
+        <div
+          ref={pickerRef}
+          className="ep-add-clip-picker ep-add-clip-picker--portal"
+          style={{
+            position: "fixed",
+            top: pickerPos.top,
+            right: pickerPos.right,
+          }}
+        >
+          <div className="ep-add-clip-picker-title">添加片段</div>
+
+          {/* 类型选择 */}
+          <div className="ep-add-clip-type-row">
+            <span className="ep-add-clip-type-label">类型:</span>
+            {availableTypes.map((t) => (
+              <button
+                key={t}
+                className={`ep-add-clip-type-btn${addType === t ? " active" : ""}`}
+                onClick={() => setAddType(t)}
+              >
+                {CLIP_TYPE_ICONS[t]} {CLIP_TYPE_LABELS[t]}
+              </button>
+            ))}
+          </div>
+
+          {/* 时长输入 */}
+          <div className="ep-add-clip-duration-row">
+            <span className="ep-add-clip-type-label">时长:</span>
+            <input
+              type="number"
+              className="ep-duration-input"
+              min={1}
+              max={120}
+              value={addDuration}
+              onChange={(e) =>
+                setAddDuration(
+                  Math.max(1, Math.min(120, Number(e.target.value) || 1)),
+                )
+              }
+            />
+            <span className="ep-add-clip-duration-unit">秒</span>
+          </div>
+
+          {/* 确认按钮 */}
+          <button
+            className="ep-add-clip-confirm-btn"
+            onClick={handleConfirmAdd}
+          >
+            添加
+          </button>
+        </div>
+      )}
     </div>
   );
 };

apps/web/src/pages/editing-planner/types.ts

@@ -0,0 +1,21 @@
+/**
+ * 片段（Clip）统一类型定义
+ * 片段 = 时间规划 + 类型标记，不绑定任何素材
+ */
+
+export type ClipType = "voice" | "pip";
+
+export interface ClipData {
+  id: string;
+  type: ClipType; // 片段类型：voice（口播）或 pip（画中画）
+  duration: number; // 时长（秒）
+  startOffset: number; // 仅 voice 类型：在口播素材中的起始时间（秒）
+  // 保留兼容字段（后端序列化需要）
+  template_segment_id?: string;
+  script_text?: string;
+  order?: number;
+  /** 配音素材 ID（voice 类型片段使用） */
+  voice_asset_id?: string;
+  /** 配音素材文件 URL（voice 类型片段使用） */
+  voice_file_url?: string;
+}

apps/web/src/pages/generate/generate.css

@@ -408,24 +408,41 @@
 }
 
 .xx-clone-avatar.ready {
-  background: linear-gradient(135deg, var(--color-secondary-500), var(--color-secondary-600));
+  background: linear-gradient(
+    135deg,
+    var(--color-secondary-500),
+    var(--color-secondary-600)
+  );
   color: var(--text-inverse);
 }
 
 .xx-clone-avatar.processing {
-  background: linear-gradient(135deg, var(--color-accent-500), var(--color-accent-600));
+  background: linear-gradient(
+    135deg,
+    var(--color-accent-500),
+    var(--color-accent-600)
+  );
   color: var(--text-inverse);
   animation: xx-clone-pulse 2s ease-in-out infinite;
 }
 
 .xx-clone-avatar.failed {
-  background: linear-gradient(135deg, var(--color-gray-400), var(--color-gray-500));
+  background: linear-gradient(
+    135deg,
+    var(--color-gray-400),
+    var(--color-gray-500)
+  );
   color: var(--text-inverse);
 }
 
 @keyframes xx-clone-pulse {
-  0%, 100% { opacity: 1; }
-  50% { opacity: 0.6; }
+  0%,
+  100% {
+    opacity: 1;
+  }
+  50% {
+    opacity: 0.6;
+  }
 }
 
 .xx-clone-info {
@@ -465,7 +482,11 @@
 
 .xx-clone-progress-bar {
   height: 100%;
-  background: linear-gradient(90deg, var(--warning-color), var(--success-color));
+  background: linear-gradient(
+    90deg,
+    var(--warning-color),
+    var(--success-color)
+  );
   border-radius: 3px;
   transition: width 0.5s ease;
 }
@@ -474,8 +495,10 @@
   width: 100%;
   background: repeating-linear-gradient(
     90deg,
-    var(--warning-color) 0%, var(--warning-color) 25%,
-    var(--success-color) 25%, var(--success-color) 50%,
+    var(--warning-color) 0%,
+    var(--warning-color) 25%,
+    var(--success-color) 25%,
+    var(--success-color) 50%,
     var(--warning-color) 50%
   );
   background-size: 60px 100%;
@@ -483,8 +506,12 @@
 }
 
 @keyframes xx-clone-progress-flow {
-  from { background-position: 0 0; }
-  to { background-position: 60px 0; }
+  from {
+    background-position: 0 0;
+  }
+  to {
+    background-position: 60px 0;
+  }
 }
 
 .xx-clone-progress-text {
@@ -518,13 +545,23 @@
 }
 
 @keyframes xx-fade {
-  0%, 100% { opacity: 1; }
-  50% { opacity: 0.7; }
+  0%,
+  100% {
+    opacity: 1;
+  }
+  50% {
+    opacity: 0.7;
+  }
 }
 
 @keyframes xx-blink {
-  0%, 100% { opacity: 1; }
-  50% { opacity: 0.3; }
+  0%,
+  100% {
+    opacity: 1;
+  }
+  50% {
+    opacity: 0.3;
+  }
 }
 
 /* ============================================================
@@ -640,7 +677,11 @@
   aspect-ratio: 9 / 16;
   max-height: 400px;
   border-radius: var(--radius-md);
-  background: linear-gradient(135deg, var(--color-gray-900), var(--color-primary-900));
+  background: linear-gradient(
+    135deg,
+    var(--color-gray-900),
+    var(--color-primary-900)
+  );
   display: grid;
   place-items: center;
   color: var(--text-inverse);
@@ -654,7 +695,11 @@
   content: "";
   position: absolute;
   inset: 0;
-  background: radial-gradient(circle at 72% 28%, rgba(255, 255, 255, 0.2), transparent 40%);
+  background: radial-gradient(
+    circle at 72% 28%,
+    rgba(255, 255, 255, 0.2),
+    transparent 40%
+  );
   pointer-events: none;
 }
 
@@ -777,7 +822,11 @@
 
 .xx-progress-bar-fill {
   height: 100%;
-  background: linear-gradient(90deg, var(--color-primary-500), var(--color-primary-600));
+  background: linear-gradient(
+    90deg,
+    var(--color-primary-500),
+    var(--color-primary-600)
+  );
   border-radius: 3px;
   transition: width 0.3s ease;
 }
@@ -834,3 +883,332 @@
     flex-direction: column;
   }
 }
+
+/* ── 存为素材弹窗 ── */
+.xx-save-modal-overlay {
+  position: fixed;
+  inset: 0;
+  background: rgba(0, 0, 0, 0.45);
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  z-index: 1000;
+  animation: xxFadeIn 0.15s ease;
+}
+
+.xx-save-modal {
+  background: var(--bg-card, #fff);
+  border-radius: var(--radius-lg, 16px);
+  width: 420px;
+  max-width: 90vw;
+  box-shadow: 0 20px 60px rgba(0, 0, 0, 0.15);
+  animation: xxSlideUp 0.2s ease;
+}
+
+.xx-save-modal-header {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  padding: 16px 20px;
+  border-bottom: 1px solid var(--border-light, #f1f5f9);
+  font-weight: 600;
+  font-size: 15px;
+  color: var(--text-primary, #0f172a);
+}
+
+.xx-save-modal-close {
+  background: none;
+  border: none;
+  cursor: pointer;
+  color: var(--text-tertiary, #94a3b8);
+  font-size: 14px;
+  padding: 4px;
+  border-radius: var(--radius-sm, 6px);
+  transition: all 0.15s;
+}
+
+.xx-save-modal-close:hover {
+  background: var(--bg-hover, #f8fafc);
+  color: var(--text-primary, #0f172a);
+}
+
+.xx-save-modal-body {
+  padding: 20px;
+}
+
+.xx-save-modal-label {
+  display: block;
+  font-size: 13px;
+  font-weight: 500;
+  color: var(--text-secondary, #475569);
+  margin-bottom: 6px;
+  margin-top: 14px;
+}
+
+.xx-save-modal-label:first-child {
+  margin-top: 0;
+}
+
+.xx-save-modal-input {
+  width: 100%;
+  padding: 8px 12px;
+  border: 1px solid var(--border-color, #e2e8f0);
+  border-radius: var(--radius-sm, 10px);
+  font-size: 14px;
+  outline: none;
+  background: var(--bg-surface, #fff);
+  color: var(--text-primary, #0f172a);
+  transition: border-color 0.15s;
+}
+
+.xx-save-modal-input:focus {
+  border-color: var(--primary-500, #6366f1);
+  box-shadow: 0 0 0 2px var(--primary-100, #e0e7ff);
+}
+
+.xx-save-modal-tags {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 6px;
+  padding: 8px;
+  border: 1px solid var(--border-color, #e2e8f0);
+  border-radius: var(--radius-sm, 10px);
+  min-height: 40px;
+  align-items: center;
+  cursor: text;
+  transition: border-color 0.15s;
+}
+
+.xx-save-modal-tags:focus-within {
+  border-color: var(--primary-500, #6366f1);
+  box-shadow: 0 0 0 2px var(--primary-100, #e0e7ff);
+}
+
+.xx-save-modal-tag {
+  display: inline-flex;
+  align-items: center;
+  gap: 4px;
+  padding: 2px 8px;
+  font-size: 12px;
+  border-radius: 12px;
+  background: var(--primary-50, #eef2ff);
+  color: var(--primary-600, #4f46e5);
+  border: 1px solid var(--primary-200, #c7d2fe);
+}
+
+.xx-save-modal-tag-remove {
+  font-size: 10px;
+  cursor: pointer;
+  opacity: 0.6;
+  transition: opacity 0.15s;
+}
+
+.xx-save-modal-tag-remove:hover {
+  opacity: 1;
+}
+
+.xx-save-modal-tag-input {
+  border: none;
+  outline: none;
+  flex: 1;
+  min-width: 100px;
+  font-size: 13px;
+  background: transparent;
+  color: var(--text-primary, #0f172a);
+}
+
+.xx-save-modal-tag-input::placeholder {
+  color: var(--text-tertiary, #94a3b8);
+}
+
+.xx-save-modal-tag-presets {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 6px;
+  margin-top: 10px;
+  padding-top: 10px;
+  border-top: 1px solid var(--border-light, #f1f5f9);
+}
+
+.xx-save-modal-tag-preset {
+  padding: 3px 10px;
+  font-size: 12px;
+  border: 1px solid var(--border-color, #e2e8f0);
+  border-radius: 12px;
+  background: var(--bg-surface, #fff);
+  color: var(--text-secondary, #475569);
+  cursor: pointer;
+  transition: all 0.15s;
+}
+
+.xx-save-modal-tag-preset:hover {
+  border-color: var(--primary-300, #a5b4fc);
+  color: var(--primary-600, #4f46e5);
+  background: var(--primary-50, #eef2ff);
+}
+
+.xx-save-modal-footer {
+  display: flex;
+  justify-content: flex-end;
+  gap: 8px;
+  padding: 14px 20px;
+  border-top: 1px solid var(--border-light, #f1f5f9);
+}
+
+@keyframes xxFadeIn {
+  from {
+    opacity: 0;
+  }
+  to {
+    opacity: 1;
+  }
+}
+
+@keyframes xxSlideUp {
+  from {
+    transform: translateY(12px);
+    opacity: 0;
+  }
+  to {
+    transform: translateY(0);
+    opacity: 1;
+  }
+}
+
+/* ── 生成数量步进器 ── */
+.xx-count-stepper {
+  display: inline-flex;
+  align-items: center;
+  gap: 6px;
+}
+
+.xx-count-stepper-btn {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  width: 28px;
+  height: 28px;
+  border: 1px solid var(--border-primary, #e2e8f0);
+  border-radius: 8px;
+  background: var(--bg-surface, #fff);
+  color: var(--text-secondary, #64748b);
+  font-size: 13px;
+  cursor: pointer;
+  transition: all 0.15s;
+}
+
+.xx-count-stepper-btn:hover:not(:disabled) {
+  border-color: var(--primary-400, #818cf8);
+  color: var(--primary-600, #4f46e5);
+  background: var(--primary-50, #eef2ff);
+}
+
+.xx-count-stepper-btn:disabled {
+  opacity: 0.35;
+  cursor: not-allowed;
+}
+
+.xx-count-stepper-value {
+  min-width: 24px;
+  text-align: center;
+  font-size: 16px;
+  font-weight: 600;
+  color: var(--text-primary, #1e293b);
+}
+
+.xx-count-stepper-hint {
+  font-size: 12px;
+  color: var(--text-tertiary, #94a3b8);
+  margin-left: 2px;
+}
+
+/* ── 素材选择模式切换 Tab ── */
+.xx-material-mode-tabs {
+  display: flex;
+  gap: 0;
+  border: 1px solid var(--border-primary, #e2e8f0);
+  border-radius: 10px;
+  overflow: hidden;
+  margin-bottom: 4px;
+}
+
+.xx-material-mode-tab {
+  flex: 1;
+  padding: 10px 16px;
+  font-size: 13px;
+  font-weight: 500;
+  border: none;
+  background: var(--bg-surface, #fff);
+  color: var(--text-secondary, #64748b);
+  cursor: pointer;
+  transition: all 0.2s ease;
+  text-align: center;
+}
+
+.xx-material-mode-tab:first-child {
+  border-right: 1px solid var(--border-primary, #e2e8f0);
+}
+
+.xx-material-mode-tab:hover {
+  background: var(--primary-50, #eef2ff);
+  color: var(--primary-600, #4f46e5);
+}
+
+.xx-material-mode-tab.active {
+  background: var(--primary-500, #6366f1);
+  color: #fff;
+  font-weight: 600;
+}
+
+/* ── 自动匹配卡片 ── */
+.xx-auto-match-card {
+  margin-top: 14px;
+  padding: 20px;
+  background: linear-gradient(135deg, #f0f4ff 0%, #faf5ff 100%);
+  border: 1px solid var(--border-primary, #e2e8f0);
+  border-radius: 14px;
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  text-align: center;
+}
+
+.xx-auto-match-icon {
+  font-size: 36px;
+  margin-bottom: 10px;
+}
+
+.xx-auto-match-body {
+  width: 100%;
+}
+
+.xx-auto-match-title {
+  font-size: 15px;
+  font-weight: 600;
+  color: var(--text-primary, #1e293b);
+  margin: 0 0 8px;
+}
+
+.xx-auto-match-desc {
+  font-size: 13px;
+  color: var(--text-secondary, #64748b);
+  line-height: 1.6;
+  margin: 0 0 14px;
+}
+
+.xx-auto-match-features {
+  display: flex;
+  flex-wrap: wrap;
+  justify-content: center;
+  gap: 8px;
+}
+
+.xx-auto-match-feature {
+  display: inline-block;
+  padding: 4px 12px;
+  font-size: 12px;
+  color: var(--primary-600, #4f46e5);
+  background: rgba(255, 255, 255, 0.8);
+  border: 1px solid var(--border-light, #f1f5f9);
+  border-radius: 20px;
+}

apps/web/src/pages/my-templates/MyTemplates.tsx

@@ -226,6 +226,7 @@ const MyTemplates: React.FC = () => {
                     >
                       {MODE_LABELS[tpl.mode as TemplateMode] || tpl.mode}
                     </Tag>
+                    <Tag color="green">用户自制</Tag>
                   </div>
 
                   <div className="mt-card-meta">

apps/web/src/pages/my-voices/MyVoices.tsx

@@ -16,6 +16,7 @@ import {
   ClockCircleOutlined,
 } from "@ant-design/icons";
 import { Button, Modal, Input, Tooltip } from "@/components/ui";
+import type { ButtonProps } from "antd";
 import PageHead from "@/components/layout/PageHead";
 import { useCloneProgress } from "@/hooks/useCloneProgress";
 import {
@@ -206,9 +207,11 @@ const MyVoices: React.FC = () => {
       if (audioRef.current) {
         audioRef.current.pause();
       }
-      // Mock: 使用 sample_url 或占位 URL
-      const url = voice.sample_url || `/mock/audio/clone-${voice.id}.mp3`;
-      const audio = new Audio(url);
+      if (!voice.sample_url) {
+        showToast("暂无试听音频", "error");
+        return;
+      }
+      const audio = new Audio(voice.sample_url);
       audioRef.current = audio;
       audio.play().catch(() => showToast("播放失败，请检查音频文件", "error"));
       audio.onended = () => setPlayingId(null);
@@ -378,7 +381,7 @@ const MyVoices: React.FC = () => {
         onOk={handleDeleteConfirm}
         okText="删除"
         cancelText="取消"
-        okButtonProps={{ danger: true } as any}
+        okButtonProps={{ danger: true } as ButtonProps}
       >
         <p>确定要删除这个克隆音色吗？删除后无法恢复。</p>
       </Modal>

apps/web/src/pages/my-voices/my-voices.css

@@ -112,19 +112,11 @@
 }
 
 .xx-mv-card--processing {
-  border-color: color-mix(
-    in srgb,
-    var(--accent-color) 30%,
-    transparent
-  );
+  border-color: color-mix(in srgb, var(--accent-color) 30%, transparent);
 }
 
 .xx-mv-card--failed {
-  border-color: color-mix(
-    in srgb,
-    var(--error-color) 25%,
-    transparent
-  );
+  border-color: color-mix(in srgb, var(--error-color) 25%, transparent);
   opacity: 0.85;
 }
 

apps/web/src/pages/products/ProductDetail.tsx

@@ -0,0 +1,491 @@
+/**
+ * 成品库 — 产品详情页
+ * 路由：/app/products/:id
+ * 展示视频播放器 + 完整元数据 + 下载/分享/删除操作
+ */
+import React, { useRef, useState, useEffect, useCallback } from "react";
+import { useParams, useNavigate } from "react-router-dom";
+import { useQuery, useMutation, useQueryClient } from "@tanstack/react-query";
+import {
+  ArrowLeftOutlined,
+  DownloadOutlined,
+  ShareAltOutlined,
+  DeleteOutlined,
+  PlayCircleFilled,
+  PauseCircleFilled,
+  SoundOutlined,
+  MutedOutlined,
+  ExpandOutlined,
+  LoadingOutlined,
+  WarningOutlined,
+} from "@ant-design/icons";
+import {
+  getProduct,
+  deleteProduct,
+  getProductDownloadUrl,
+  type ProductItem,
+} from "../../api/products";
+import { Button } from "../../components/ui";
+import "./products.css";
+
+/* ============================================================
+ * 工具函数
+ * ============================================================ */
+
+/** 格式化时长（秒 → "MM:SS"） */
+const formatDuration = (seconds: number): string => {
+  if (!seconds || seconds <= 0) return "00:00";
+  const m = Math.floor(seconds / 60);
+  const s = Math.floor(seconds % 60);
+  return `${String(m).padStart(2, "0")}:${String(s).padStart(2, "0")}`;
+};
+
+/** 格式化文件大小（MB） */
+const formatFileSize = (mb: number): string => {
+  if (!mb || mb <= 0) return "-";
+  if (mb < 1024) return `${mb.toFixed(1)} MB`;
+  return `${(mb / 1024).toFixed(2)} GB`;
+};
+
+/** 格式化日期 */
+const formatDate = (dateStr: string): string => {
+  if (!dateStr) return "-";
+  const d = new Date(dateStr);
+  return d.toLocaleDateString("zh-CN", {
+    year: "numeric",
+    month: "2-digit",
+    day: "2-digit",
+    hour: "2-digit",
+    minute: "2-digit",
+  });
+};
+
+/** 状态标签 */
+const STATUS_MAP: Record<string, { label: string; color: string }> = {
+  completed: { label: "已完成", color: "#10b981" },
+  processing: { label: "处理中", color: "#6366f1" },
+  pending: { label: "待处理", color: "#f59e0b" },
+  failed: { label: "失败", color: "#ef4444" },
+};
+
+/* ============================================================
+ * 主组件
+ * ============================================================ */
+const ProductDetail: React.FC = () => {
+  const { id } = useParams<{ id: string }>();
+  const navigate = useNavigate();
+  const queryClient = useQueryClient();
+
+  /* ── 获取产品详情 ── */
+  const {
+    data: product,
+    isLoading,
+    isError,
+    error,
+  } = useQuery<ProductItem, Error>({
+    queryKey: ["product", id],
+    queryFn: () => getProduct(id!),
+    enabled: !!id,
+    staleTime: 10_000,
+  });
+
+  /* ── 删除 mutation ── */
+  const deleteMutation = useMutation({
+    mutationFn: () => deleteProduct(id!),
+    onSuccess: () => {
+      queryClient.invalidateQueries({ queryKey: ["products"] });
+      navigate("/app/products");
+    },
+  });
+
+  /* ── 视频播放器状态 ── */
+  const videoRef = useRef<HTMLVideoElement>(null);
+  const progressRef = useRef<HTMLDivElement>(null);
+  const hideTimerRef = useRef<ReturnType<typeof setTimeout>>();
+
+  const [isPlaying, setIsPlaying] = useState(false);
+  const [currentTime, setCurrentTime] = useState(0);
+  const [duration, setDuration] = useState(0);
+  const [buffered, setBuffered] = useState(0);
+  const [volume, setVolume] = useState(1);
+  const [isMuted, setIsMuted] = useState(false);
+  const [showControls, setShowControls] = useState(true);
+  const [isFullscreen, setIsFullscreen] = useState(false);
+  const containerRef = useRef<HTMLDivElement>(null);
+
+  /* ── 自动隐藏控制条 ── */
+  const resetHideTimer = useCallback(() => {
+    setShowControls(true);
+    if (hideTimerRef.current) clearTimeout(hideTimerRef.current);
+    if (isPlaying) {
+      hideTimerRef.current = setTimeout(() => setShowControls(false), 3000);
+    }
+  }, [isPlaying]);
+
+  /* ── 播放控制 ── */
+  const togglePlay = useCallback(() => {
+    const v = videoRef.current;
+    if (!v) return;
+    if (v.paused) {
+      v.play().catch(() => {});
+    } else {
+      v.pause();
+    }
+  }, []);
+
+  const handleSeek = useCallback(
+    (e: React.MouseEvent<HTMLDivElement>) => {
+      const v = videoRef.current;
+      const bar = progressRef.current;
+      if (!v || !bar || !duration) return;
+      const rect = bar.getBoundingClientRect();
+      const ratio = Math.max(
+        0,
+        Math.min(1, (e.clientX - rect.left) / rect.width),
+      );
+      v.currentTime = ratio * duration;
+    },
+    [duration],
+  );
+
+  const handleVolumeChange = useCallback(
+    (e: React.ChangeEvent<HTMLInputElement>) => {
+      const v = videoRef.current;
+      const val = parseFloat(e.target.value);
+      if (v) v.volume = val;
+      setVolume(val);
+      setIsMuted(val === 0);
+    },
+    [],
+  );
+
+  const toggleMute = useCallback(() => {
+    const v = videoRef.current;
+    if (!v) return;
+    if (isMuted) {
+      v.muted = false;
+      v.volume = volume || 1;
+      setIsMuted(false);
+    } else {
+      v.muted = true;
+      setIsMuted(true);
+    }
+  }, [isMuted, volume]);
+
+  const toggleFullscreen = useCallback(() => {
+    const el = containerRef.current;
+    if (!el) return;
+    if (!document.fullscreenElement) {
+      el.requestFullscreen?.().catch(() => {});
+    } else {
+      document.exitFullscreen?.().catch(() => {});
+    }
+  }, []);
+
+  /* ── 视频事件监听 ── */
+  useEffect(() => {
+    const v = videoRef.current;
+    if (!v) return;
+
+    const onPlay = () => setIsPlaying(true);
+    const onPause = () => setIsPlaying(false);
+    const onTimeUpdate = () => setCurrentTime(v.currentTime);
+    const onLoadedMetadata = () => setDuration(v.duration);
+    const onProgress = () => {
+      if (v.buffered.length > 0) {
+        setBuffered(v.buffered.end(v.buffered.length - 1));
+      }
+    };
+    const onEnded = () => setIsPlaying(false);
+    const onFSChange = () => setIsFullscreen(!!document.fullscreenElement);
+
+    v.addEventListener("play", onPlay);
+    v.addEventListener("pause", onPause);
+    v.addEventListener("timeupdate", onTimeUpdate);
+    v.addEventListener("loadedmetadata", onLoadedMetadata);
+    v.addEventListener("progress", onProgress);
+    v.addEventListener("ended", onEnded);
+    document.addEventListener("fullscreenchange", onFSChange);
+
+    return () => {
+      v.removeEventListener("play", onPlay);
+      v.removeEventListener("pause", onPause);
+      v.removeEventListener("timeupdate", onTimeUpdate);
+      v.removeEventListener("loadedmetadata", onLoadedMetadata);
+      v.removeEventListener("progress", onProgress);
+      v.removeEventListener("ended", onEnded);
+      document.removeEventListener("fullscreenchange", onFSChange);
+    };
+  }, []);
+
+  /* 播放时自动隐藏/显示控制条 */
+  useEffect(() => {
+    resetHideTimer();
+    return () => {
+      if (hideTimerRef.current) clearTimeout(hideTimerRef.current);
+    };
+  }, [isPlaying, resetHideTimer]);
+
+  /* ── 下载 ── */
+  const handleDownload = async () => {
+    if (!product || product.status !== "completed") return;
+    try {
+      const { url } = await getProductDownloadUrl(product.id);
+      const a = document.createElement("a");
+      a.href = url;
+      a.download = "";
+      a.click();
+    } catch {
+      // message.error handled by caller
+    }
+  };
+
+  /* ── 分享 ── */
+  const handleShare = () => {
+    if (!product) return;
+    const link = `${window.location.origin}/app/products/${product.id}`;
+    navigator.clipboard?.writeText(link).then(
+      () => {},
+      () => {},
+    );
+  };
+
+  /* ── 删除 ── */
+  const handleDelete = () => {
+    if (!id) return;
+    deleteMutation.mutate();
+  };
+
+  /* ── 加载状态 ── */
+  if (isLoading) {
+    return (
+      <div className="xx-page xx-product-detail-page">
+        <div className="xx-detail-loading">
+          <LoadingOutlined
+            style={{ fontSize: 32, color: "var(--primary-color)" }}
+            spin
+          />
+          <p style={{ color: "var(--text-secondary)", marginTop: 16 }}>
+            加载中…
+          </p>
+        </div>
+      </div>
+    );
+  }
+
+  /* ── 错误状态 ── */
+  if (isError || !product) {
+    return (
+      <div className="xx-page xx-product-detail-page">
+        <div className="xx-detail-error">
+          <WarningOutlined
+            style={{ fontSize: 48, color: "var(--error-color)" }}
+          />
+          <h3>加载失败</h3>
+          <p>{error?.message || "无法获取产品信息"}</p>
+          <Button
+            buttonType="ghost"
+            buttonSize="md"
+            onClick={() => navigate("/app/products")}
+          >
+            <ArrowLeftOutlined /> 返回成品库
+          </Button>
+        </div>
+      </div>
+    );
+  }
+
+  const statusInfo = STATUS_MAP[product.status] || {
+    label: product.status,
+    color: "#94a3b8",
+  };
+  const progress = Math.round((currentTime / (duration || 1)) * 100);
+  const bufferedPct = Math.round((buffered / (duration || 1)) * 100);
+
+  return (
+    <div className="xx-page xx-product-detail-page">
+      {/* ── 顶部导航 ── */}
+      <div className="xx-detail-header">
+        <button
+          className="xx-detail-back-btn"
+          onClick={() => navigate("/app/products")}
+        >
+          <ArrowLeftOutlined /> 返回成品库
+        </button>
+        <div className="xx-detail-actions">
+          <Button
+            buttonType="ghost"
+            buttonSize="sm"
+            icon={<DownloadOutlined />}
+            onClick={handleDownload}
+            disabled={product.status !== "completed"}
+          >
+            下载
+          </Button>
+          <Button
+            buttonType="ghost"
+            buttonSize="sm"
+            icon={<ShareAltOutlined />}
+            onClick={handleShare}
+            disabled={product.status !== "completed"}
+          >
+            分享
+          </Button>
+          <Button
+            buttonType="danger"
+            buttonSize="sm"
+            icon={<DeleteOutlined />}
+            onClick={handleDelete}
+            loading={deleteMutation.isPending}
+          >
+            删除
+          </Button>
+        </div>
+      </div>
+
+      {/* ── 主体内容 ── */}
+      <div className="xx-detail-body">
+        {/* 视频播放器 */}
+        <div
+          ref={containerRef}
+          className={`xx-detail-player ${isFullscreen ? "is-fullscreen" : ""}`}
+          onMouseMove={resetHideTimer}
+          onClick={togglePlay}
+        >
+          {product.video_url ? (
+            <video
+              ref={videoRef}
+              src={product.video_url}
+              poster={product.thumbnail_url || undefined}
+              playsInline
+              preload="metadata"
+            />
+          ) : (
+            <div className="xx-detail-player-empty">
+              {product.thumbnail_url ? (
+                <img src={product.thumbnail_url} alt={product.title} />
+              ) : (
+                <div className="xx-detail-player-placeholder">
+                  <PlayCircleFilled
+                    style={{ fontSize: 64, color: "var(--text-tertiary)" }}
+                  />
+                  <p>视频暂不可用</p>
+                </div>
+              )}
+            </div>
+          )}
+
+          {/* 控制条 */}
+          <div
+            className={`xx-detail-player-controls ${showControls || !isPlaying ? "visible" : ""}`}
+            onClick={(e) => e.stopPropagation()}
+          >
+            {/* 进度条 */}
+            <div
+              className="xx-dp-progress"
+              ref={progressRef}
+              onClick={handleSeek}
+            >
+              <div className="xx-dp-progress-track">
+                <div
+                  className="xx-dp-progress-buffered"
+                  style={{ width: `${bufferedPct}%` }}
+                />
+                <div
+                  className="xx-dp-progress-played"
+                  style={{ width: `${progress}%` }}
+                />
+              </div>
+            </div>
+
+            <div className="xx-dp-bar">
+              {/* 左：播放/暂停 */}
+              <button className="xx-dp-btn" onClick={togglePlay}>
+                {isPlaying ? <PauseCircleFilled /> : <PlayCircleFilled />}
+              </button>
+
+              {/* 时间 */}
+              <span className="xx-dp-time">
+                {formatDuration(currentTime)} / {formatDuration(duration)}
+              </span>
+
+              <div style={{ flex: 1 }} />
+
+              {/* 音量 */}
+              <div className="xx-dp-volume">
+                <button className="xx-dp-btn" onClick={toggleMute}>
+                  {isMuted ? <MutedOutlined /> : <SoundOutlined />}
+                </button>
+                <input
+                  type="range"
+                  min={0}
+                  max={1}
+                  step={0.05}
+                  value={isMuted ? 0 : volume}
+                  onChange={handleVolumeChange}
+                  className="xx-dp-volume-slider"
+                />
+              </div>
+
+              {/* 全屏 */}
+              <button className="xx-dp-btn" onClick={toggleFullscreen}>
+                <ExpandOutlined />
+              </button>
+            </div>
+          </div>
+        </div>
+
+        {/* 产品信息 */}
+        <div className="xx-detail-info">
+          <h1 className="xx-detail-title">{product.title}</h1>
+
+          <div className="xx-detail-meta-grid">
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">状态</span>
+              <span
+                className="xx-detail-meta-value"
+                style={{ color: statusInfo.color }}
+              >
+                {statusInfo.label}
+              </span>
+            </div>
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">时长</span>
+              <span className="xx-detail-meta-value">
+                {formatDuration(product.duration_seconds ?? 0)}
+              </span>
+            </div>
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">分辨率</span>
+              <span className="xx-detail-meta-value">
+                {product.resolution || "-"}
+              </span>
+            </div>
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">文件大小</span>
+              <span className="xx-detail-meta-value">
+                {formatFileSize(product.file_size ?? 0)}
+              </span>
+            </div>
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">查重率</span>
+              <span className="xx-detail-meta-value">
+                {(product.duplicate_rate ?? 0) > 0
+                  ? `${(product.duplicate_rate ?? 0).toFixed(1)}%`
+                  : "-"}
+              </span>
+            </div>
+            <div className="xx-detail-meta-item">
+              <span className="xx-detail-meta-label">创建时间</span>
+              <span className="xx-detail-meta-value">
+                {formatDate(product.created_at ?? "")}
+              </span>
+            </div>
+          </div>
+        </div>
+      </div>
+    </div>
+  );
+};
+
+export default ProductDetail;

apps/web/src/pages/products/ProductLibrary.tsx

@@ -11,6 +11,7 @@ import React, {
   useCallback,
 } from "react";
 import { useQuery, useMutation, useQueryClient } from "@tanstack/react-query";
+import { useNavigate } from "react-router-dom";
 import { message, Popconfirm } from "antd";
 import {
   SearchOutlined,
@@ -23,6 +24,7 @@ import {
   CloseOutlined,
   CheckOutlined,
   CloudUploadOutlined,
+  EyeOutlined,
 } from "@ant-design/icons";
 import { Button, Input, Select } from "@/components/ui";
 import {
@@ -311,7 +313,8 @@ const VideoPlayer: React.FC<{
   onClose: () => void;
   onDownload: (product: ProductItem) => void;
   onShare: (product: ProductItem) => void;
-}> = ({ product, onClose, onDownload, onShare }) => {
+  onViewDetail: (product: ProductItem) => void;
+}> = ({ product, onClose, onDownload, onShare, onViewDetail }) => {
   const videoRef = useRef<HTMLVideoElement>(null);
   const progressRef = useRef<HTMLDivElement>(null);
   const [isPlaying, setIsPlaying] = useState(false);
@@ -480,6 +483,14 @@ const VideoPlayer: React.FC<{
           >
             分享
           </Button>
+          <Button
+            buttonType="ghost"
+            buttonSize="sm"
+            icon={<EyeOutlined />}
+            onClick={() => onViewDetail(product)}
+          >
+            查看详情
+          </Button>
           <div style={{ flex: 1 }} />
           <Button buttonType="primary" buttonSize="sm" onClick={onClose}>
             关闭
@@ -495,6 +506,7 @@ const VideoPlayer: React.FC<{
  * ============================================================ */
 const ProductLibrary: React.FC = () => {
   const queryClient = useQueryClient();
+  const navigate = useNavigate();
 
   /* ── 获取成品列表 ── */
   const {
@@ -645,6 +657,12 @@ const ProductLibrary: React.FC = () => {
     );
   };
 
+  /* 查看详情 — 跳转到产品详情页 */
+  const handleViewDetail = (product: ProductItem) => {
+    setPlayingProduct(null); // 关闭播放器
+    navigate(`/app/products/${product.id}`);
+  };
+
   /* 删除 */
   const handleDelete = (id: string) => {
     deleteMutation.mutate(id);
@@ -719,11 +737,39 @@ const ProductLibrary: React.FC = () => {
 
   // ── Error 状态 ──
   if (isError) {
+    console.error("[ProductLibrary] 加载失败:", error);
+    const errorMsg = error?.message || "加载失败";
+    // 404 视为空数据（API 尚未就绪或无数据）
+    const is404 = errorMsg.includes("404") || errorMsg.includes("Not Found");
+    if (is404) {
+      return (
+        <div className="xx-products-page">
+          <div className="xx-products-header">
+            <h2>
+              <VideoCameraOutlined /> 成片库
+            </h2>
+          </div>
+          <div className="xx-products-empty">
+            <div className="xx-products-empty-icon">🎬</div>
+            <p>暂无成片数据</p>
+            <p
+              style={{
+                fontSize: 12,
+                color: "var(--text-tertiary)",
+                marginTop: 4,
+              }}
+            >
+              完成视频生成后，成片将自动保存到这里
+            </p>
+          </div>
+        </div>
+      );
+    }
     return (
       <div className="xx-products-page">
         <div className="xx-products-empty">
           <div className="xx-products-empty-icon">❌</div>
-          <p>{error?.message || "加载失败"}</p>
+          <p>{errorMsg || "加载失败，请稍后重试"}</p>
           <Button
             buttonType="primary"
             buttonSize="sm"
@@ -912,6 +958,7 @@ const ProductLibrary: React.FC = () => {
           onClose={() => setPlayingProduct(null)}
           onDownload={handleDownload}
           onShare={handleShare}
+          onViewDetail={handleViewDetail}
         />
       )}
     </div>

apps/web/src/pages/products/products.css

@@ -631,6 +631,327 @@
   opacity: 0.5;
 }
 
+/* ============================================================
+   产品详情页
+   ============================================================ */
+.xx-product-detail-page {
+  padding: var(--space-lg);
+}
+
+.xx-detail-loading,
+.xx-detail-error {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  justify-content: center;
+  min-height: 60vh;
+  gap: var(--space-sm);
+}
+
+.xx-detail-error h3 {
+  font-size: 20px;
+  font-weight: var(--font-weight-bold);
+  color: var(--text-primary);
+  margin: var(--space-md) 0 0;
+}
+
+.xx-detail-error p {
+  color: var(--text-secondary);
+  font-size: 14px;
+  margin: 0 0 var(--space-lg);
+}
+
+/* 顶部导航 */
+.xx-detail-header {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  margin-bottom: var(--space-lg);
+  padding: var(--space-md) var(--space-lg);
+  background: rgba(255, 255, 255, 0.94);
+  border: 1px solid rgba(226, 232, 240, 0.95);
+  border-radius: var(--radius-xl);
+  box-shadow: var(--shadow-xs);
+}
+
+.xx-detail-back-btn {
+  display: inline-flex;
+  align-items: center;
+  gap: 6px;
+  padding: 8px 16px;
+  font-size: 14px;
+  font-weight: 600;
+  color: var(--text-secondary);
+  background: transparent;
+  border: 1px solid var(--border-color);
+  border-radius: var(--radius-sm);
+  cursor: pointer;
+  transition: all var(--transition-fast);
+}
+
+.xx-detail-back-btn:hover {
+  color: var(--primary-color);
+  border-color: var(--primary-color);
+  background: var(--primary-soft);
+}
+
+.xx-detail-actions {
+  display: flex;
+  gap: var(--space-sm);
+}
+
+/* 主体 */
+.xx-detail-body {
+  display: grid;
+  grid-template-columns: 1fr 360px;
+  gap: var(--space-lg);
+  align-items: start;
+}
+
+@media (max-width: 1100px) {
+  .xx-detail-body {
+    grid-template-columns: 1fr;
+  }
+}
+
+/* 播放器 */
+.xx-detail-player {
+  position: relative;
+  width: 100%;
+  aspect-ratio: 16 / 9;
+  background: #000;
+  border-radius: var(--radius-lg);
+  overflow: hidden;
+  cursor: pointer;
+  box-shadow: var(--shadow-md);
+}
+
+.xx-detail-player video {
+  width: 100%;
+  height: 100%;
+  object-fit: contain;
+}
+
+.xx-detail-player.is-fullscreen {
+  border-radius: 0;
+}
+
+.xx-detail-player-empty {
+  width: 100%;
+  height: 100%;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+}
+
+.xx-detail-player-empty img {
+  max-width: 100%;
+  max-height: 100%;
+  object-fit: contain;
+}
+
+.xx-detail-player-placeholder {
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: var(--space-sm);
+  color: var(--text-tertiary);
+}
+
+.xx-detail-player-placeholder p {
+  font-size: 14px;
+  margin: 0;
+}
+
+/* 播放器控制条 */
+.xx-detail-player-controls {
+  position: absolute;
+  left: 0;
+  right: 0;
+  bottom: 0;
+  background: linear-gradient(transparent, rgba(0, 0, 0, 0.7));
+  padding: var(--space-lg) var(--space-md) var(--space-md);
+  opacity: 0;
+  transition: opacity var(--transition-base);
+  pointer-events: none;
+}
+
+.xx-detail-player-controls.visible {
+  opacity: 1;
+  pointer-events: auto;
+}
+
+/* 进度条 */
+.xx-dp-progress {
+  width: 100%;
+  height: 20px;
+  display: flex;
+  align-items: center;
+  cursor: pointer;
+  margin-bottom: var(--space-xs);
+}
+
+.xx-dp-progress-track {
+  position: relative;
+  width: 100%;
+  height: 4px;
+  background: rgba(255, 255, 255, 0.25);
+  border-radius: 2px;
+  overflow: hidden;
+  transition: height var(--transition-fast);
+}
+
+.xx-dp-progress:hover .xx-dp-progress-track {
+  height: 6px;
+}
+
+.xx-dp-progress-buffered {
+  position: absolute;
+  top: 0;
+  left: 0;
+  height: 100%;
+  background: rgba(255, 255, 255, 0.35);
+  border-radius: 2px;
+}
+
+.xx-dp-progress-played {
+  position: absolute;
+  top: 0;
+  left: 0;
+  height: 100%;
+  background: var(--primary-color);
+  border-radius: 2px;
+}
+
+/* 控制栏 */
+.xx-dp-bar {
+  display: flex;
+  align-items: center;
+  gap: var(--space-sm);
+}
+
+.xx-dp-btn {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  width: 32px;
+  height: 32px;
+  border: none;
+  background: transparent;
+  color: #fff;
+  font-size: 18px;
+  cursor: pointer;
+  border-radius: var(--radius-xs);
+  transition: background var(--transition-fast);
+}
+
+.xx-dp-btn:hover {
+  background: rgba(255, 255, 255, 0.15);
+}
+
+.xx-dp-time {
+  font-size: 12px;
+  color: rgba(255, 255, 255, 0.8);
+  font-variant-numeric: tabular-nums;
+  white-space: nowrap;
+}
+
+.xx-dp-volume {
+  display: flex;
+  align-items: center;
+  gap: 4px;
+}
+
+.xx-dp-volume-slider {
+  width: 72px;
+  height: 4px;
+  appearance: none;
+  background: rgba(255, 255, 255, 0.3);
+  border-radius: 2px;
+  outline: none;
+  cursor: pointer;
+}
+
+.xx-dp-volume-slider::-webkit-slider-thumb {
+  appearance: none;
+  width: 12px;
+  height: 12px;
+  border-radius: 50%;
+  background: #fff;
+  cursor: pointer;
+}
+
+.xx-dp-volume-slider::-moz-range-thumb {
+  width: 12px;
+  height: 12px;
+  border-radius: 50%;
+  background: #fff;
+  border: none;
+  cursor: pointer;
+}
+
+/* 产品信息面板 */
+.xx-detail-info {
+  background: rgba(255, 255, 255, 0.94);
+  border: 1px solid rgba(226, 232, 240, 0.95);
+  border-radius: var(--radius-xl);
+  padding: var(--space-xl);
+  box-shadow: var(--shadow-card);
+}
+
+.xx-detail-title {
+  font-size: 22px;
+  font-weight: var(--font-weight-bold);
+  color: var(--text-primary);
+  margin: 0 0 var(--space-lg);
+  line-height: var(--line-height-tight);
+}
+
+.xx-detail-meta-grid {
+  display: grid;
+  grid-template-columns: 1fr 1fr;
+  gap: var(--space-md);
+}
+
+.xx-detail-meta-item {
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+}
+
+.xx-detail-meta-label {
+  font-size: 12px;
+  font-weight: 500;
+  color: var(--text-tertiary);
+}
+
+.xx-detail-meta-value {
+  font-size: 14px;
+  font-weight: 600;
+  color: var(--text-primary);
+}
+
+/* 详情页响应式 */
+@media (max-width: 768px) {
+  .xx-product-detail-page {
+    padding: var(--space-md);
+  }
+
+  .xx-detail-header {
+    flex-direction: column;
+    gap: var(--space-sm);
+    align-items: stretch;
+  }
+
+  .xx-detail-actions {
+    justify-content: flex-end;
+  }
+
+  .xx-detail-meta-grid {
+    grid-template-columns: 1fr;
+  }
+}
+
 /* ============================================================
    响应式
    ============================================================ */

apps/web/src/pages/templates/TemplateLibrary.tsx

@@ -91,7 +91,7 @@ const mapTemplateItemToEditTemplate = (item: TemplateItem): EditTemplate => ({
   id: item.id,
   name: item.name,
   type: inferTemplateType(item.category),
-  description: item.description,
+  description: item.description ?? "",
   usageCount: 0,
   isFavorite: item.is_favorite ?? false,
   thumbnailGradient: gradientForCategory(item.category),
@@ -370,9 +370,10 @@ const TemplateCard: React.FC<TemplateCardProps> = ({
           className="xx-template-thumb-bg"
           style={{ background: template.thumbnailGradient }}
         >
-          {template.description.slice(0, 80)}...
+          {(template.description ?? "").slice(0, 80)}...
         </div>
         <div className="xx-template-thumb-overlay" />
+        <div className="xx-template-thumb-name">{template.name}</div>
         <div className="xx-template-preview-hint">点击预览</div>
         <button
           className={`xx-template-fav-btn${isFavorite ? " is-favorite" : ""}`}
@@ -386,7 +387,6 @@ const TemplateCard: React.FC<TemplateCardProps> = ({
       {/* 信息区 */}
       <div className="xx-template-info">
         <div className="xx-template-info-top">
-          <h4 className="xx-template-name">{template.name}</h4>
           <span
             className="xx-template-category-pill"
             style={{
@@ -397,7 +397,7 @@ const TemplateCard: React.FC<TemplateCardProps> = ({
             {template.type}
           </span>
         </div>
-        <p className="xx-template-desc">{template.description}</p>
+        <p className="xx-template-desc">{template.description ?? ""}</p>
         <div className="xx-template-meta">
           <span className="xx-template-usage">
             已使用 {template.usageCount} 次
@@ -483,7 +483,9 @@ const TemplateLibrary: React.FC = () => {
       const matchSearch =
         !searchText ||
         t.name.toLowerCase().includes(searchText.toLowerCase()) ||
-        t.description.toLowerCase().includes(searchText.toLowerCase()) ||
+        (t.description ?? "")
+          .toLowerCase()
+          .includes(searchText.toLowerCase()) ||
         t.tags.some((tag) =>
           tag.toLowerCase().includes(searchText.toLowerCase()),
         );

apps/web/src/pages/templates/templates.css

@@ -233,6 +233,24 @@
   pointer-events: none;
 }
 
+/* 缩略图底部名称 */
+.xx-template-thumb-name {
+  position: absolute;
+  bottom: 0;
+  left: 0;
+  right: 0;
+  padding: 24px 14px 10px;
+  background: linear-gradient(0deg, rgba(0, 0, 0, 0.55) 0%, transparent 100%);
+  color: #fff;
+  font-size: 14px;
+  font-weight: 600;
+  white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  z-index: 1;
+  pointer-events: none;
+}
+
 /* 预览提示（hover 显示） */
 .xx-template-preview-hint {
   position: absolute;

apps/web/src/pages/voice-clone/voice-clone.css

@@ -36,8 +36,7 @@
 
 .vc-card:hover {
   border-color: var(--primary-light);
-  box-shadow: 0 4px 20px
-    color-mix(in srgb, var(--primary-color) 8%);
+  box-shadow: 0 4px 20px color-mix(in srgb, var(--primary-color) 8%);
 }
 
 /* 右上角操作按钮 */
@@ -103,7 +102,7 @@
   width: 52px;
   height: 52px;
   border-radius: 50%;
-  background: linear-gradient(135deg, var(--accent-color)));
+  background: linear-gradient(135deg, var(--accent-color));
   display: grid;
   place-items: center;
   font-size: 22px;
@@ -282,7 +281,7 @@
   font-weight: 500;
   background: var(--bg-card);
   border: 1px solid var(--line);
-  box-shadow: 0 4px 12px var(--shadow-sm));
+  box-shadow: 0 4px 12px var(--shadow-sm);
   animation: vc-toast-in 0.25s ease-out;
   display: flex;
   align-items: center;
@@ -318,7 +317,7 @@
   z-index: 1000;
   display: grid;
   place-items: center;
-  background: var(--overlay-bg));
+  background: var(--overlay-bg);
   animation: vc-fade-in 0.2s;
 }
 
@@ -337,7 +336,7 @@
   padding: 24px;
   width: 400px;
   max-width: calc(100vw - 40px);
-  box-shadow: 0 20px 60px var(--shadow-lg));
+  box-shadow: 0 20px 60px var(--shadow-lg);
   animation: vc-scale-in 0.2s ease-out;
 }
 

apps/web/src/router/index.tsx

@@ -26,6 +26,7 @@ const ProtectedRoute = ({ children }: { children: React.ReactNode }) => {
 };
 
 /** 首页路由组件：已登录跳 dashboard，未登录显示落地页 */
+// eslint-disable-next-line react-refresh/only-export-components
 const HomeRoute: React.FC = () => {
   const isAuthenticated = useAuthStore((state) => state.isAuthenticated);
   const hasAccessToken = Boolean(localStorage.getItem("access_token"));
@@ -127,6 +128,13 @@ export const router = createBrowserRouter([
             Component: m.default,
           })),
       },
+      {
+        path: "products/:id",
+        lazy: () =>
+          import("@/pages/products/ProductDetail").then((m) => ({
+            Component: m.default,
+          })),
+      },
       {
         path: "editing-planner",
         lazy: () =>
@@ -148,6 +156,13 @@ export const router = createBrowserRouter([
             Component: m.default,
           })),
       },
+      {
+        path: "voice-materials",
+        lazy: () =>
+          import("@/pages/voice-materials/VoiceMaterialLibrary").then((m) => ({
+            Component: m.default,
+          })),
+      },
       {
         path: "my-voices",
         lazy: () =>

apps/web/src/store/authStore.ts

@@ -75,6 +75,8 @@ export const useAuthStore = create<AuthState>()(
       partialize: (state) => ({
         user: state.user,
         isAuthenticated: state.isAuthenticated,
+        accessToken: state.accessToken,
+        refreshToken: state.refreshToken,
       }),
     },
   ),

apps/worker/video_processing/dedup.py

@@ -206,6 +206,68 @@ class VideoDeduplicator:
 
         return None
 
+    def check_batch_duplicate(
+        self,
+        fingerprint: VideoFingerprint,
+        batch_id: str,
+        current_video_id: str,
+        session: Session,
+    ) -> Optional[dict]:
+        """检查视频是否与同批次内其他视频重复。
+
+        逻辑与 check_duplicate 一致（MD5 + pHash），但搜索范围限定为同 batch_id 的视频。
+
+        Args:
+            fingerprint: 待检测视频的指纹
+            batch_id: 批次 ID
+            current_video_id: 当前视频 ID（排除自身）
+            session: 数据库会话
+
+        Returns:
+            重复信息字典，或 None 表示未找到重复
+        """
+        video_repo = SQLAlchemyGeneratedVideoRepository(session)
+        batch_videos = video_repo.list_by_batch(batch_id)
+
+        for existing in batch_videos:
+            if existing.id == current_video_id:
+                continue
+            if not existing.video_fingerprint:
+                continue
+
+            ef = existing.video_fingerprint
+
+            if fingerprint.md5 == ef.get("md5"):
+                return {
+                    "duplicate": True,
+                    "duplicate_of": existing.id,
+                    "reason": "batch_exact_md5_match",
+                    "similarity": 1.0,
+                }
+
+            existing_phashes = ef.get("keyframe_phashes", [])
+            if not existing_phashes:
+                continue
+
+            min_distances = []
+            for phash in fingerprint.keyframe_phashes:
+                distances = [hamming_distance(phash, ep) for ep in existing_phashes]
+                min_distances.append(min(distances))
+            avg_distance = sum(min_distances) / len(min_distances) if min_distances else 100
+
+            if avg_distance >= self.PHASH_THRESHOLD:
+                continue
+
+            phash_similarity = 1.0 - (avg_distance / 64)
+            return {
+                "duplicate": True,
+                "duplicate_of": existing.id,
+                "reason": "batch_phash_similar",
+                "similarity": phash_similarity,
+            }
+
+        return None
+
     @staticmethod
     def _average_histogram_similarity(histograms_a: list[list[float]], histograms_b: list[list[float]]) -> float:
         """

apps/worker/worker_app/tasks/__init__.py

@@ -41,6 +41,10 @@ def __getattr__(name: str):
         from .tts_synthesis import process_tts_synthesis
 
         return process_tts_synthesis
+    elif name == "process_tts_segment_synthesis":
+        from .tts_synthesis import process_tts_segment_synthesis
+
+        return process_tts_segment_synthesis
     elif name == "run_ai_recommend":
         from .ai_tasks import run_ai_recommend
 
@@ -62,6 +66,7 @@ __all__ = [
     "extract_background_task",
     "process_voice_clone",
     "process_tts_synthesis",
+    "process_tts_segment_synthesis",
     "run_ai_recommend",
     "run_generate_cover",
 ]

apps/worker/worker_app/tasks/generation.py

@@ -139,14 +139,16 @@ def _download_library_assets(
     asset_library_id: str,
     temp_path: Path,
     video_extensions: tuple = (".mp4", ".mov", ".avi", ".mkv", ".webm"),
+    asset_ids: list[str] | None = None,
 ) -> list[str]:
     """
-    从素材库下载所有视频素材
+    从素材库下载视频素材
 
     Args:
         asset_library_id: 素材库 ID
         temp_path: 临时目录路径
         video_extensions: 支持的视频扩展名
+        asset_ids: 指定素材 ID 列表，为空则下载全部 ready 视频素材
 
     Returns:
         下载成功的视频文件路径列表
@@ -161,16 +163,15 @@ def _download_library_assets(
 
         try:
             # 查询素材库中的视频素材
-            assets = (
-                session.query(AssetModel)
-                .filter(
-                    AssetModel.asset_library_id == asset_library_id,
-                    AssetModel.status == "ready",
-                    AssetModel.file_type.in_(["video", "video/mp4", "video/quicktime"]),
-                )
-                .order_by(AssetModel.created_at)
-                .all()
+            query = session.query(AssetModel).filter(
+                AssetModel.asset_library_id == asset_library_id,
+                AssetModel.status == "ready",
+                AssetModel.file_type.in_(["video", "video/mp4", "video/quicktime"]),
             )
+            # 如果指定了 asset_ids，则只下载这些素材
+            if asset_ids:
+                query = query.filter(AssetModel.id.in_(asset_ids))
+            assets = query.order_by(AssetModel.created_at).all()
 
             if not assets:
                 logger.info(f"No video assets found in library {asset_library_id}")
@@ -259,6 +260,8 @@ def generate_video(self, task_id: str) -> dict:
         asset_library_id = gen_task.asset_library_id
         voice_library_id = gen_task.voice_library_id or ""
         mode = gen_task.strategy_id or "one_take"
+        task_asset_ids = list(gen_task.asset_ids or [])
+        batch_id = getattr(gen_task, "batch_id", "") or ""
     finally:
         session.close()
 
@@ -275,8 +278,8 @@ def generate_video(self, task_id: str) -> dict:
             temp_path = Path(temp_dir)
             output_path = temp_path / output_name
 
-            # 从素材库下载视频素材
-            downloaded_videos = _download_library_assets(asset_library_id, temp_path)
+            # 从素材库下载视频素材（如果任务指定了 asset_ids 则只下载这些）
+            downloaded_videos = _download_library_assets(asset_library_id, temp_path, asset_ids=task_asset_ids or None)
 
             audio_path = None
             if voice_library_id:
@@ -297,6 +300,32 @@ def generate_video(self, task_id: str) -> dict:
             file_size = output_path.stat().st_size
             duration = _probe_duration(output_path)
 
+            # 上传到 OSS
+            bucket = _oss_bucket()
+            if bucket:
+                try:
+                    bucket.put_object_from_file(storage_key, str(output_path))
+                except Exception as oss_err:
+                    logger.warning(f"OSS upload failed: {oss_err}")
+
+            # 构建视频 URL
+            if bucket:
+                file_url = f"{PUBLIC_API_BASE_URL}/{storage_key}"
+            else:
+                file_url = f"{GENERATED_FILES_URL_PREFIX}/{task_id}/{output_name}"
+
+            # 创建 GeneratedVideo 记录 + 查重
+            _create_video_record_and_dedup(
+                task_id=task_id,
+                project_id=project_id,
+                batch_id=batch_id,
+                file_url=file_url,
+                file_size=file_size,
+                duration=duration,
+                video_path=str(output_path),
+                mode=editing_mode.value,
+            )
+
             return {
                 "status": "completed",
                 "task_id": task_id,
@@ -314,3 +343,85 @@ def generate_video(self, task_id: str) -> dict:
             "task_id": task_id,
             "error": str(error),
         }
+
+
+def _create_video_record_and_dedup(
+    *,
+    task_id: str,
+    project_id: str,
+    batch_id: str,
+    file_url: str,
+    file_size: int,
+    duration: float,
+    video_path: str,
+    mode: str,
+) -> None:
+    """创建 GeneratedVideo 记录，计算指纹并执行查重（历史 + 批次）。"""
+    from uuid import uuid4
+
+    from video_processing.dedup import VideoDeduplicator
+    from worker_app.db import SessionLocal
+
+    from packages.adapters.sqlalchemy_impl.generated_video_repository import (
+        SQLAlchemyGeneratedVideoRepository,
+    )
+    from packages.domain import GeneratedVideo
+
+    session = SessionLocal()
+    try:
+        video_id = uuid4().hex
+        generated_video = GeneratedVideo(
+            id=video_id,
+            project_id=project_id,
+            generation_task_id=task_id,
+            name=f"generated-{task_id[:8]}.mp4",
+            file_url=file_url,
+            file_size=file_size,
+            duration=duration,
+            width=OUTPUT_WIDTH,
+            height=OUTPUT_HEIGHT,
+            fps=OUTPUT_FPS,
+            status="completed",
+            generation_params={"mode": mode},
+        )
+
+        video_repo = SQLAlchemyGeneratedVideoRepository(session)
+        video_repo.create(generated_video)
+
+        # 计算视频指纹
+        deduplicator = VideoDeduplicator()
+        try:
+            fingerprint = deduplicator.compute_fingerprint(video_path)
+        except Exception as fp_err:
+            logger.warning(f"Fingerprint computation failed for {video_id}: {fp_err}")
+            session.commit()
+            return
+
+        generated_video.video_fingerprint = fingerprint.to_dict()
+
+        # (a) 历史成片查重
+        duplicate_result = deduplicator.check_duplicate(fingerprint, project_id, session)
+
+        # (b) 批次内查重（仅当有 batch_id 时）
+        if not duplicate_result and batch_id:
+            duplicate_result = deduplicator.check_batch_duplicate(fingerprint, batch_id, video_id, session)
+
+        if duplicate_result:
+            generated_video.is_duplicate = True
+            generated_video.duplicate_of = duplicate_result["duplicate_of"]
+            logger.info(
+                f"Duplicate detected: {video_id} -> {duplicate_result['duplicate_of']} "
+                f"(reason={duplicate_result['reason']}, similarity={duplicate_result['similarity']:.3f})"
+            )
+        else:
+            generated_video.is_duplicate = False
+            generated_video.duplicate_of = None
+
+        video_repo.update(generated_video)
+        session.commit()
+        logger.info(f"GeneratedVideo record created: {video_id} (task={task_id}, dup={generated_video.is_duplicate})")
+    except Exception as e:
+        logger.error(f"Failed to create video record / dedup for task {task_id}: {e}")
+        session.rollback()
+    finally:
+        session.close()

apps/worker/worker_app/tasks/ingest.py

@@ -179,6 +179,7 @@ def ingest_asset(job_id: str) -> dict:
             width=int(metadata.get("width", 0)),
             height=int(metadata.get("height", 0)),
             status=AssetStatus.READY,
+            file_hash=job.file_hash,
         )
         asset_repo.create(asset)
 

apps/worker/worker_app/tasks/tts_synthesis.py

@@ -104,3 +104,77 @@ def process_tts_synthesis(self: Task, job_id: str) -> dict:
     finally:
         if session is not None:
             session.close()
+
+
+@celery_app.task(bind=True, max_retries=2, name="worker.process_tts_segment_synthesis")
+def process_tts_segment_synthesis(self: Task, job_id: str) -> dict:
+    """分段合成轮询任务 — 轮询多个 CosyVoice 子任务并合并音频。
+
+    与 process_tts_synthesis 类似，但超时更长（300s），
+    因为分段任务需要等待所有子任务完成。
+    """
+    session = None
+    try:
+        session = SessionLocal()
+        repo = SQLAlchemyTTSJobRepository(session)
+        workflow = TTSWorkflowService(
+            repository=repo,
+            cosyvoice_service=CosyVoiceService(),
+        )
+
+        updated_job = workflow.poll_and_process_synthesis(job_id, timeout=300)
+        session.commit()
+
+        logger.info(f"TTS segment synthesis completed: job_id={job_id}, " f"audio_url={updated_job.output_audio_url}")
+        return {
+            "ok": True,
+            "job_id": job_id,
+            "audio_url": updated_job.output_audio_url,
+        }
+
+    except Retry:
+        raise
+
+    except CosyVoiceTimeoutError as e:
+        logger.warning(f"TTS segment synthesis timeout for {job_id}: {e}")
+        if session is not None:
+            session.rollback()
+        raise self.retry(exc=e, countdown=60)
+
+    except CosyVoiceError as e:
+        logger.error(f"TTS segment synthesis failed for {job_id}: {e}")
+        if session is not None:
+            session.rollback()
+        try:
+            if session is not None:
+                job = repo.get(job_id)
+                if job is not None:
+                    job.mark_failed(str(e))
+                    repo.update(job)
+                    session.commit()
+        except Exception as inner_e:
+            logger.error(f"Failed to mark job as failed: {inner_e}")
+            if session is not None:
+                session.rollback()
+        return {"ok": False, "job_id": job_id, "error": str(e)}
+
+    except Exception as e:
+        logger.error(f"TTS segment synthesis unexpected error for {job_id}: {e}")
+        if session is not None:
+            session.rollback()
+        try:
+            if session is not None:
+                job = repo.get(job_id)
+                if job is not None:
+                    job.mark_failed(str(e))
+                    repo.update(job)
+                    session.commit()
+        except Exception as inner_e:
+            logger.error(f"Failed to mark job as failed: {inner_e}")
+            if session is not None:
+                session.rollback()
+        return {"ok": False, "job_id": job_id, "error": str(e)}
+
+    finally:
+        if session is not None:
+            session.close()

docs/schema-metadata-snapshot.json

@@ -95,6 +95,39 @@
         "id"
       ]
     },
+    "asset_tags": {
+      "columns": [
+        {
+          "index": false,
+          "name": "asset_id",
+          "nullable": false,
+          "primary_key": true,
+          "type": "VARCHAR(36)",
+          "unique": false
+        },
+        {
+          "index": false,
+          "name": "tag_id",
+          "nullable": false,
+          "primary_key": true,
+          "type": "VARCHAR(36)",
+          "unique": false
+        },
+        {
+          "index": false,
+          "name": "created_at",
+          "nullable": false,
+          "primary_key": false,
+          "type": "DATETIME",
+          "unique": false
+        }
+      ],
+      "indexes": [],
+      "primary_key": [
+        "asset_id",
+        "tag_id"
+      ]
+    },
     "assets": {
       "columns": [
         {
@@ -241,6 +274,14 @@
           "type": "VARCHAR(36)",
           "unique": false
         },
+        {
+          "index": true,
+          "name": "file_hash",
+          "nullable": true,
+          "primary_key": false,
+          "type": "VARCHAR(64)",
+          "unique": false
+        },
         {
           "index": false,
           "name": "metadata",
@@ -288,6 +329,13 @@
           "name": "ix_assets_created_at",
           "unique": false
         },
+        {
+          "columns": [
+            "file_hash"
+          ],
+          "name": "ix_assets_file_hash",
+          "unique": false
+        },
         {
           "columns": [
             "file_type"
@@ -1469,6 +1517,22 @@
           "type": "VARCHAR(32)",
           "unique": false
         },
+        {
+          "index": false,
+          "name": "asset_select_mode",
+          "nullable": false,
+          "primary_key": false,
+          "type": "VARCHAR(20)",
+          "unique": false
+        },
+        {
+          "index": true,
+          "name": "batch_id",
+          "nullable": false,
+          "primary_key": false,
+          "type": "VARCHAR(32)",
+          "unique": false
+        },
         {
           "index": false,
           "name": "metadata",
@@ -1494,6 +1558,13 @@
           "name": "ix_generation_tasks_asset_library_id",
           "unique": false
         },
+        {
+          "columns": [
+            "batch_id"
+          ],
+          "name": "ix_generation_tasks_batch_id",
+          "unique": false
+        },
         {
           "columns": [
             "created_by_user_id"
@@ -1599,6 +1670,14 @@
           "type": "VARCHAR(32)",
           "unique": false
         },
+        {
+          "index": true,
+          "name": "file_hash",
+          "nullable": true,
+          "primary_key": false,
+          "type": "VARCHAR(64)",
+          "unique": false
+        },
         {
           "index": false,
           "name": "created_at",
@@ -1617,6 +1696,13 @@
         }
       ],
       "indexes": [
+        {
+          "columns": [
+            "file_hash"
+          ],
+          "name": "ix_ingest_jobs_file_hash",
+          "unique": false
+        },
         {
           "columns": [
             "library_id"
@@ -2056,6 +2142,54 @@
         "id"
       ]
     },
+    "tags": {
+      "columns": [
+        {
+          "index": false,
+          "name": "id",
+          "nullable": false,
+          "primary_key": true,
+          "type": "VARCHAR(36)",
+          "unique": false
+        },
+        {
+          "index": true,
+          "name": "user_id",
+          "nullable": false,
+          "primary_key": false,
+          "type": "VARCHAR(36)",
+          "unique": false
+        },
+        {
+          "index": false,
+          "name": "name",
+          "nullable": false,
+          "primary_key": false,
+          "type": "VARCHAR(100)",
+          "unique": false
+        },
+        {
+          "index": false,
+          "name": "created_at",
+          "nullable": false,
+          "primary_key": false,
+          "type": "DATETIME",
+          "unique": false
+        }
+      ],
+      "indexes": [
+        {
+          "columns": [
+            "user_id"
+          ],
+          "name": "ix_tags_user_id",
+          "unique": false
+        }
+      ],
+      "primary_key": [
+        "id"
+      ]
+    },
     "template_categories": {
       "columns": [
         {

infra/docker/deploy-production-registry.sh

@@ -0,0 +1,240 @@
+#!/bin/sh
+set -eu
+
+# ============================================
+# Production 部署脚本 - Registry 方式
+# 用法：IMAGE_TAG=<version> REGISTRY_TOKEN=<token> sh deploy-production-registry.sh
+# ============================================
+
+IMAGE_TAG="${IMAGE_TAG:-}"
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
+REGISTRY_TOKEN="${REGISTRY_TOKEN:-}"
+
+ENV_FILE="${ENV_FILE:-/var/lib/xiaoxia-saas-production/.env}"
+GENERATED_DIR="${GENERATED_DIR:-/var/lib/xiaoxia-saas-production/generated}"
+LEGACY_ASSETS_DIR="${LEGACY_ASSETS_DIR:-/var/lib/xiaoxia-saas-production/legacy-assets}"
+
+if [ -z "$IMAGE_TAG" ]; then
+  echo "ERROR: IMAGE_TAG is required"
+  exit 1
+fi
+
+test -f "$ENV_FILE"
+mkdir -p "$GENERATED_DIR"
+mkdir -p "$LEGACY_ASSETS_DIR"
+
+# ---- 登录 Registry ----
+if [ -n "$REGISTRY_TOKEN" ]; then
+  echo "Logging in to registry: $REGISTRY"
+  REGISTRY_HOST=$(echo "$REGISTRY" | cut -d/ -f1)
+  printf %s "$REGISTRY_TOKEN" | docker login "$REGISTRY_HOST" -u "$REGISTRY_USER" --password-stdin 2>/dev/null || {
+    echo "WARN: docker login failed, will try to pull anyway"
+  }
+fi
+
+# ---- Pull 三镜像 ----
+REGISTRY_API="${REGISTRY}/xiaoxia-saas-api:${IMAGE_TAG}"
+REGISTRY_WORKER="${REGISTRY}/xiaoxia-saas-worker:${IMAGE_TAG}"
+REGISTRY_WEB="${REGISTRY}/xiaoxia-saas-web:${IMAGE_TAG}"
+
+LOCAL_API="xiaoxia-saas-api:${IMAGE_TAG}"
+LOCAL_WORKER="xiaoxia-saas-worker:${IMAGE_TAG}"
+LOCAL_WEB="xiaoxia-saas-web:${IMAGE_TAG}"
+
+echo "Pulling API image..."
+docker pull "$REGISTRY_API"
+echo "Pulling Worker image..."
+docker pull "$REGISTRY_WORKER"
+echo "Pulling Web image..."
+docker pull "$REGISTRY_WEB"
+
+# ---- Re-tag 成本地名 ----
+docker tag "$REGISTRY_API" "$LOCAL_API"
+docker tag "$REGISTRY_WORKER" "$LOCAL_WORKER"
+docker tag "$REGISTRY_WEB" "$LOCAL_WEB"
+echo "All images pulled and tagged."
+
+# ---- 备份旧版 assets（部署期间缓存用户不 404） ----
+echo "Backing up legacy assets from current web container..."
+if docker inspect xiaoxia-web-production >/dev/null 2>&1; then
+  _tmpdir="/tmp/legacy-assets-$$"
+  rm -rf "$_tmpdir"
+  mkdir -p "$_tmpdir"
+  docker cp xiaoxia-web-production:/usr/share/nginx/html/assets/. "$_tmpdir/" 2>/dev/null || true
+  # 合并到 LEGACY_ASSETS_DIR（保留所有历史版本的 assets）
+  if [ -d "$_tmpdir" ] && [ "$(ls -A "$_tmpdir" 2>/dev/null)" ]; then
+    cp -an "$_tmpdir"/. "$LEGACY_ASSETS_DIR"/ 2>/dev/null || true
+    echo "Legacy assets backed up: $(ls "$_tmpdir" | wc -l) files"
+  fi
+  rm -rf "$_tmpdir"
+else
+  echo "No existing web container, skipping legacy assets backup"
+fi
+
+# 清理超过 7 天的旧 assets 文件（避免无限增长）
+if [ -d "$LEGACY_ASSETS_DIR" ]; then
+  find "$LEGACY_ASSETS_DIR" -type f -mtime +7 -delete 2>/dev/null || true
+  echo "Legacy assets cleanup done (retain 7 days)"
+fi
+
+# ---- 确保基础设施容器在运行 ----
+echo "Checking infrastructure containers..."
+for c in xiaoxia-postgres-production xiaoxia-redis-production; do
+  if ! docker inspect "$c" >/dev/null 2>&1; then
+    echo "ERROR: Required container not found: $c"
+    exit 1
+  fi
+  state=$(docker inspect -f '{{.State.Status}}' "$c")
+  if [ "$state" != "running" ]; then
+    echo "ERROR: Container not running: $c ($state)"
+    exit 1
+  fi
+done
+
+# ---- 确保生产网络存在 ----
+docker network create xiaoxia-net-production 2>/dev/null || true
+
+# ---- 执行数据库 Migration ----
+echo "Running database migrations..."
+docker run --rm \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-production \
+  -e APP_ENV=production \
+  "$LOCAL_API" sh -c "cd /app && alembic upgrade head"
+echo "Migrations completed."
+
+# ---- 停止旧容器 ----
+echo "Stopping old containers..."
+docker rm -f xiaoxia-api-production 2>/dev/null || true
+docker rm -f xiaoxia-worker-production 2>/dev/null || true
+docker rm -f xiaoxia-web-production 2>/dev/null || true
+
+# ---- 日志配置（所有容器共用） ----
+LOG_OPTS="--log-driver json-file --log-opt max-size=50m --log-opt max-file=3"
+
+# ---- 启动 API ----
+echo "Starting API container..."
+docker run -d \
+  --name xiaoxia-api-production \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-production \
+  -p 127.0.0.1:8001:8000 \
+  -e APP_ENV=production \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --cpus 2 \
+  --memory 2g \
+  --health-cmd "python -c \"import urllib.request; urllib.request.urlopen('http://localhost:8000/health', timeout=5)\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 40s \
+  $LOG_OPTS \
+  "$LOCAL_API"
+
+# ---- 启动 Worker ----
+echo "Starting Worker container..."
+docker run -d \
+  --name xiaoxia-worker-production \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-production \
+  -e APP_ENV=production \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e WORKER_CONCURRENCY=1 \
+  -e WORKER_MAX_TASKS_PER_CHILD=100 \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --cpus 2 \
+  --memory 2g \
+  --health-cmd "sh -c \"grep -q celery /proc/1/cmdline || exit 1\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 30s \
+  $LOG_OPTS \
+  "$LOCAL_WORKER"
+
+# ---- 启动 Web ----
+# Legacy assets 挂载到 /usr/share/nginx/html/assets-legacy/assets/
+# nginx 配置中 assets location 有 fallback 逻辑
+LEGACY_VOLUME=""
+if [ -d "$LEGACY_ASSETS_DIR" ] && [ "$(ls -A "$LEGACY_ASSETS_DIR" 2>/dev/null)" ]; then
+  LEGACY_VOLUME="-v ${LEGACY_ASSETS_DIR}:/usr/share/nginx/html/assets-legacy/assets:ro"
+  echo "Web container: legacy assets mounted (fallback)"
+else
+  echo "Web container: no legacy assets to mount"
+fi
+
+echo "Starting Web container..."
+docker run -d \
+  --name xiaoxia-web-production \
+  --network xiaoxia-net-production \
+  -p 127.0.0.1:3002:80 \
+  --restart unless-stopped \
+  --cpus 0.5 \
+  --memory 512m \
+  $LEGACY_VOLUME \
+  --health-cmd "wget --spider -q http://127.0.0.1:80" \
+  --health-interval 30s \
+  --health-timeout 5s \
+  --health-retries 3 \
+  $LOG_OPTS \
+  "$LOCAL_WEB"
+
+# ---- 等待 API 健康 ----
+echo "Waiting for API to become healthy..."
+i=0
+while [ "$i" -lt 40 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:8001/health >/dev/null 2>&1; then
+    echo "API is healthy!"
+    break
+  fi
+  i=$((i + 1))
+  echo "  Waiting... ($i/40)"
+  sleep 3
+done
+
+if [ "$i" -ge 40 ]; then
+  echo "ERROR: API did not become healthy within 120s"
+  docker logs --tail 50 xiaoxia-api-production
+  exit 1
+fi
+
+# ---- 等待 Web 健康 ----
+echo "Waiting for Web to become healthy..."
+i=0
+while [ "$i" -lt 15 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:3002/ >/dev/null 2>&1; then
+    echo "Web is healthy!"
+    break
+  fi
+  i=$((i + 1))
+  echo "  Waiting... ($i/15)"
+  sleep 2
+done
+
+if [ "$i" -ge 15 ]; then
+  echo "ERROR: Web did not become healthy within 30s"
+  docker logs --tail 30 xiaoxia-web-production
+  exit 1
+fi
+
+# ---- 清理旧镜像 ----
+echo "Cleaning up old images..."
+docker image prune -af --filter "until=168h" 2>/dev/null || true
+docker builder prune -af --filter "until=168h" 2>/dev/null || true
+
+echo ""
+echo "=== Production deployment complete ==="
+echo "API:      http://127.0.0.1:8001"
+echo "Web:      http://127.0.0.1:3002"
+echo "Version:  $IMAGE_TAG"
+docker ps --format "table {{.Names}}\t{{.Status}}\t{{.Image}}" | grep production

infra/docker/deploy-production.sh

@@ -6,25 +6,19 @@ ROOT_DIR="$HOST_PREFIX/var/lib/xiaoxia-saas-production/repo"
 COMPOSE_DIR="$ROOT_DIR/infra/docker"
 ENV_FILE="$HOST_PREFIX/var/lib/xiaoxia-saas-production/.env"
 RELEASE_VERSION="${RELEASE_VERSION:-}"
-RUNTIME_IMAGE_TAR="${RUNTIME_IMAGE_TAR:-}"
-WEB_IMAGE_TAR="${WEB_IMAGE_TAR:-}"
 
-if [ -n "$RELEASE_VERSION" ] && [ -z "$RUNTIME_IMAGE_TAR" ]; then
-  RUNTIME_IMAGE_TAR="$HOST_PREFIX/var/lib/xiaoxia-saas-production/runtime-images-$RELEASE_VERSION.tar"
-fi
-if [ -n "$RELEASE_VERSION" ] && [ -z "$WEB_IMAGE_TAR" ]; then
-  WEB_IMAGE_TAR="$HOST_PREFIX/var/lib/xiaoxia-saas-production/web-$RELEASE_VERSION.tar"
-fi
+# Registry 配置
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
+REGISTRY_TOKEN="${REGISTRY_TOKEN:-}"
 
 ensure_container_running() {
   name="$1"
   if ! docker inspect "$name" >/dev/null 2>&1; then
     echo "Required infrastructure container not found: $name"
-    echo "Run infra/docker/infra-production.yml first before deploying production applications."
     exit 1
   fi
-
-  state="$(docker inspect -f '{{.State.Status}}' "$name")"
+  state="$(docker inspect -f "{{.State.Status}}" "$name")"
   if [ "$state" != "running" ]; then
     echo "Required infrastructure container is not running: $name ($state)"
     exit 1
@@ -39,10 +33,10 @@ fi
 cd "$ROOT_DIR"
 cp "$ENV_FILE" "$ROOT_DIR/.env"
 mkdir -p "$ROOT_DIR/apps/web/public" "$HOST_PREFIX/var/lib/xiaoxia-saas-production/generated"
-[ -f "$ROOT_DIR/apps/web/public/.keep" ] || printf 'placeholder' > "$ROOT_DIR/apps/web/public/.keep"
+[ -f "$ROOT_DIR/apps/web/public/.keep" ] || printf "placeholder" > "$ROOT_DIR/apps/web/public/.keep"
 if [ ! -f "$ROOT_DIR/apps/web/dist/index.html" ]; then
   echo "Missing prebuilt web artifact: $ROOT_DIR/apps/web/dist/index.html"
-  echo "Production deploy must not build frontend assets on the server. Run npm run build before packaging the release artifact."
+  echo "Production deploy must not build frontend assets on the server."
   exit 1
 fi
 
@@ -50,23 +44,33 @@ ensure_container_running xiaoxia-postgres-production
 ensure_container_running xiaoxia-redis-production
 
 cd "$COMPOSE_DIR"
+
 if [ -n "$RELEASE_VERSION" ]; then
-  if [ ! -f "$RUNTIME_IMAGE_TAR" ]; then
-    echo "Missing production runtime image artifact: $RUNTIME_IMAGE_TAR"
-    echo "Build it on a dedicated build host with scripts/build_release_images.sh $RELEASE_VERSION, then upload it before production deploy."
-    exit 1
-  fi
-  docker load -i "$RUNTIME_IMAGE_TAR"
-  if [ ! -f "$WEB_IMAGE_TAR" ]; then
-    echo "Missing production web image artifact: $WEB_IMAGE_TAR"
-    echo "Build it on a dedicated build host, then upload it before production deploy."
-    exit 1
-  fi
-  docker load -i "$WEB_IMAGE_TAR"
   export API_IMAGE="xiaoxia-saas-api:$RELEASE_VERSION"
   export WORKER_IMAGE="xiaoxia-saas-worker:$RELEASE_VERSION"
-  export WEB_IMAGE="${WEB_IMAGE:-xiaoxia-saas-web:$RELEASE_VERSION}"
+  export WEB_IMAGE="xiaoxia-saas-web:$RELEASE_VERSION"
   export APP_VERSION="$RELEASE_VERSION"
+
+  # ---- 从 Registry Pull 镜像 ----
+  echo "Pulling images from registry: $REGISTRY"
+  if [ -n "$REGISTRY_TOKEN" ]; then
+    printf "%s" "$REGISTRY_TOKEN" | docker login git.xiaoxiajianji.com -u "$REGISTRY_USER" --password-stdin 2>/dev/null || true
+  fi
+
+  REGISTRY_API="${REGISTRY}/xiaoxia-saas-api:$RELEASE_VERSION"
+  REGISTRY_WORKER="${REGISTRY}/xiaoxia-saas-worker:$RELEASE_VERSION"
+  REGISTRY_WEB="${REGISTRY}/xiaoxia-saas-web:$RELEASE_VERSION"
+
+  docker pull "$REGISTRY_API"
+  docker pull "$REGISTRY_WORKER"
+  docker pull "$REGISTRY_WEB"
+
+  # Re-tag 为本地镜像名，保持 compose 兼容
+  docker tag "$REGISTRY_API" "$API_IMAGE"
+  docker tag "$REGISTRY_WORKER" "$WORKER_IMAGE"
+  docker tag "$REGISTRY_WEB" "$WEB_IMAGE"
+
+  echo "All images pulled and tagged."
 fi
 
 export DOCKER_BUILDKIT=0
@@ -76,7 +80,6 @@ export ENV=production
 export WEB_DOCKERFILE=infra/docker/web-artifact.Dockerfile
 export WEB_NGINX_CONF=infra/docker/nginx-production.conf
 
-# Ensure isolated production network exists
 docker network create xiaoxia-net-production 2>/dev/null || true
 export WORKER_CONCURRENCY="${WORKER_CONCURRENCY:-1}"
 export WORKER_MAX_TASKS_PER_CHILD="${WORKER_MAX_TASKS_PER_CHILD:-100}"
@@ -86,24 +89,23 @@ if [ "${ALLOW_PRODUCTION_BUILDS:-false}" = "true" ]; then
   docker compose --env-file "$ENV_FILE" build --pull=false worker
   docker compose --env-file "$ENV_FILE" build --pull=false web
 else
-  echo "Skipping production API/worker/web image builds. Set ALLOW_PRODUCTION_BUILDS=true only on a dedicated build host."
+  echo "Skipping production image builds (using prebuilt images from registry)."
   docker image inspect "${API_IMAGE:-xiaoxia-saas-api:dev}" >/dev/null
   docker image inspect "${WORKER_IMAGE:-xiaoxia-saas-worker:dev}" >/dev/null
   docker image inspect "${WEB_IMAGE:-xiaoxia-saas-web:dev}" >/dev/null
 fi
-docker compose --env-file "$ENV_FILE" run --rm --no-deps api sh -c '
+
+docker compose --env-file "$ENV_FILE" run --rm --no-deps api sh -c "
   cd /app &&
   python /app/scripts/validate_release_env.py --from-environ --strict-external &&
   alembic upgrade head
-'
+"
 docker compose --env-file "$ENV_FILE" up -d api worker
-# Recreate web after API so nginx resolves the current API container IP.
-# Docker's embedded DNS is resolved by nginx at startup for this static upstream.
 docker compose --env-file "$ENV_FILE" up -d --force-recreate web
 docker compose --env-file "$ENV_FILE" ps
 
 if [ -n "$RELEASE_VERSION" ] && [ -d "$HOST_PREFIX/etc/cron.d" ]; then
-  cat > "$HOST_PREFIX/etc/cron.d/xiaoxia-production-resource-check" <<EOF
+  cat > "$HOST_PREFIX/etc/cron.d/xiaoxia-production-resource-check" << EOF
 */5 * * * * root cd /var/lib/xiaoxia-saas-production/repo && EXPECTED_VERSION=$RELEASE_VERSION sh scripts/production_resource_check.sh >/var/log/xiaoxia-resource-check.log 2>&1
 EOF
 fi

infra/docker/deploy-staging-registry.sh

@@ -0,0 +1,178 @@
+#!/bin/sh
+set -eu
+
+# ============================================
+# Staging 部署脚本 - Registry 方式
+# 用法：IMAGE_TAG=<sha|version> REGISTRY_TOKEN=<token> sh deploy-staging.sh
+# ============================================
+
+IMAGE_TAG="${IMAGE_TAG:-}"
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
+REGISTRY_TOKEN="${REGISTRY_TOKEN:-}"
+
+ENV_FILE="${ENV_FILE:-/var/lib/xiaoxia-saas-staging/.env}"
+COMPOSE_DIR="${COMPOSE_DIR:-/var/lib/xiaoxia-saas-staging/repo/infra/docker}"
+GENERATED_DIR="${GENERATED_DIR:-/var/lib/xiaoxia-saas-staging/generated}"
+
+if [ -z "$IMAGE_TAG" ]; then
+  echo "ERROR: IMAGE_TAG is required"
+  exit 1
+fi
+
+test -f "$ENV_FILE"
+mkdir -p "$GENERATED_DIR"
+
+# ---- 登录 Registry ----
+if [ -n "$REGISTRY_TOKEN" ]; then
+  echo "Logging in to registry: $REGISTRY"
+  printf %s "$REGISTRY_TOKEN" | docker login "$(echo $REGISTRY | cut -d/ -f1)" -u "$REGISTRY_USER" --password-stdin 2>/dev/null || {
+    echo "WARN: docker login failed, will try to pull anyway"
+  }
+fi
+
+# ---- Pull 三镜像 ----
+REGISTRY_API="${REGISTRY}/xiaoxia-saas-api:${IMAGE_TAG}"
+REGISTRY_WORKER="${REGISTRY}/xiaoxia-saas-worker:${IMAGE_TAG}"
+REGISTRY_WEB="${REGISTRY}/xiaoxia-saas-web:${IMAGE_TAG}"
+
+LOCAL_API="${REGISTRY}/xiaoxia-saas-api:staging"
+LOCAL_WORKER="${REGISTRY}/xiaoxia-saas-worker:staging"
+LOCAL_WEB="${REGISTRY}/xiaoxia-saas-web:staging"
+
+echo "Pulling API image..."
+docker pull "$REGISTRY_API"
+echo "Pulling Worker image..."
+docker pull "$REGISTRY_WORKER"
+echo "Pulling Web image..."
+docker pull "$REGISTRY_WEB"
+
+# ---- Re-tag 成本地名 ----
+docker tag "$REGISTRY_API" "$LOCAL_API"
+docker tag "$REGISTRY_WORKER" "$LOCAL_WORKER"
+docker tag "$REGISTRY_WEB" "$LOCAL_WEB"
+echo "All images pulled and tagged."
+
+# ---- 确保基础设施容器在运行 ----
+for c in xiaoxia-postgres-staging xiaoxia-redis-staging; do
+  if ! docker inspect "$c" >/dev/null 2>&1; then
+    echo "ERROR: Required container not found: $c"
+    exit 1
+  fi
+  state=$(docker inspect -f {{.State.Status}} "$c")
+  if [ "$state" != "running" ]; then
+    echo "ERROR: Container not running: $c ($state)"
+    exit 1
+  fi
+done
+
+# ---- 确保 staging 网络存在 ----
+docker network create xiaoxia-net-staging 2>/dev/null || true
+
+# ---- 执行数据库 Migration ----
+echo "Running database migrations..."
+docker run --rm --env-file "$ENV_FILE" --network xiaoxia-net-staging "$LOCAL_API" sh -c "cd /app && alembic upgrade head"
+echo "Migrations completed."
+
+# ---- 停止旧容器 ----
+docker rm -f xiaoxia-api-staging 2>/dev/null || true
+docker rm -f xiaoxia-worker-staging 2>/dev/null || true
+docker rm -f xiaoxia-web-staging 2>/dev/null || true
+
+# ---- 启动 API ----
+echo "Starting API container..."
+docker run -d \
+  --name xiaoxia-api-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:8000:8000 \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --label com.centurylinklabs.watchtower.enable=true \
+  --health-cmd "python -c \"import urllib.request; urllib.request.urlopen('http://localhost:8000/health', timeout=5)\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 40s \
+  "$LOCAL_API"
+
+# ---- 启动 Worker ----
+echo "Starting Worker container..."
+docker run -d \
+  --name xiaoxia-worker-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e WORKER_CONCURRENCY=1 \
+  -e WORKER_MAX_TASKS_PER_CHILD=100 \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --label com.centurylinklabs.watchtower.enable=true \
+  --health-cmd "sh -c \"grep -q celery /proc/1/cmdline || exit 1\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 30s \
+  "$LOCAL_WORKER"
+
+# ---- 启动 Web ----
+# Web 镜像默认打包 production nginx.conf，staging 需要挂载 staging 配置
+NGINX_CONF="${NGINX_CONF:-${COMPOSE_DIR}/nginx-staging.conf}"
+if [ ! -f "$NGINX_CONF" ]; then
+  echo "WARN: nginx config not found at $NGINX_CONF, using image default"
+  NGINX_VOLUME=""
+else
+  NGINX_VOLUME="-v ${NGINX_CONF}:/etc/nginx/conf.d/default.conf:ro"
+fi
+
+echo "Starting Web container..."
+docker run -d \
+  --name xiaoxia-web-staging \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:3001:80 \
+  --restart unless-stopped \
+  --label com.centurylinklabs.watchtower.enable=true \
+  $NGINX_VOLUME \
+  --health-cmd "wget --spider -q http://127.0.0.1:80" \
+  --health-interval 30s \
+  --health-timeout 5s \
+  --health-retries 3 \
+  "$LOCAL_WEB"
+
+# ---- 等待 API 健康 ----
+echo "Waiting for API to become healthy..."
+i=0
+while [ "$i" -lt 30 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:8000/health >/dev/null 2>&1; then
+    echo "API is healthy!"
+    break
+  fi
+  i=$((i + 1))
+  echo "  Waiting... ($i/30)"
+  sleep 2
+done
+
+if [ "$i" -ge 30 ]; then
+  echo "ERROR: API did not become healthy within 60s"
+  docker logs --tail 30 xiaoxia-api-staging
+  exit 1
+fi
+
+# ---- 清理旧镜像 ----
+docker image prune -af --filter "until=72h" 2>/dev/null || true
+
+echo ""
+echo "=== Staging deployment complete ==="
+echo "API:      http://127.0.0.1:8000"
+echo "Web:      http://127.0.0.1:3001"
+echo "Version:  $IMAGE_TAG"
+docker ps --format "table {{.Names}}\t{{.Status}}" | grep staging
+
+# Watchtower auto-update: 容器加com.centurylinklabs.watchtower.enable=true标签，用:staging tag启动

infra/docker/nginx-production.conf

@@ -39,6 +39,15 @@ server {
         alias /app/generated/;
     }
 
+    # Assets with legacy fallback (higher priority than generic static regex)
+    # 部署期间，缓存了旧版 index.html 的用户会请求旧版带 hash 的 assets 文件
+    # 先在当前镜像中找，找不到去 legacy-assets 目录找（从旧版本容器中备份的）
+    location ^~ /assets/ {
+        expires 1y;
+        add_header Cache-Control "public, immutable";
+        try_files $uri /assets-legacy$uri =404;
+    }
+
     # Cache static assets
     location ~* \.(js|css|png|jpg|jpeg|gif|ico|svg|woff|woff2|ttf|eot)$ {
         expires 1y;

infra/docker/web.Dockerfile

@@ -5,7 +5,9 @@ ARG VITE_API_URL=https://saas-api.xiaoxiajianji.com
 ENV VITE_API_URL=$VITE_API_URL
 COPY apps/web/package.json apps/web/package-lock.json ./apps/web/
 WORKDIR /app/apps/web
-RUN npm config set registry https://registry.npmmirror.com \
+RUN \
+    --mount=type=cache,target=/root/.npm \
+    npm config set registry https://registry.npmmirror.com \
     && npm ci
 COPY apps/web/ ./
 RUN npm run build

packages/adapters/in_memory/asset_repository.py

@@ -26,6 +26,13 @@ class InMemoryAssetRepository:
         """Alias for list_by_library to match the port interface."""
         return self.list_by_library(library_id)
 
+    def find_by_library_and_file_type(self, library_id: str, file_type: str) -> list[Asset]:
+        return [
+            asset
+            for asset in self._assets.values()
+            if asset.library_id == library_id and asset.mime_type and asset.mime_type.startswith(file_type)
+        ]
+
     def update(self, asset: Asset) -> Asset:
         self._assets[asset.id] = asset
         return asset
@@ -35,3 +42,50 @@ class InMemoryAssetRepository:
             del self._assets[asset_id]
             return True
         return False
+
+    def batch_delete(self, asset_ids: list[str]) -> int:
+        """批量删除素材，返回实际删除数量。"""
+        count = 0
+        for aid in asset_ids:
+            if aid in self._assets:
+                del self._assets[aid]
+                count += 1
+        return count
+
+    def find_by_project(
+        self,
+        project_id: str,
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Asset]:
+        items = [a for a in self._assets.values() if a.project_id == project_id]
+        return items[skip : skip + limit]
+
+    def find_by_id(self, asset_id: str) -> Asset | None:
+        return self._assets.get(asset_id)
+
+    def find_by_tag_ids(
+        self,
+        tag_ids: list[str],
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Asset]:
+        """查找包含所有指定标签的素材。"""
+        if not tag_ids:
+            return []
+        tag_set = set(tag_ids)
+        items = [a for a in self._assets.values() if tag_set.issubset(set(a.tag_ids))]
+        return items[skip : skip + limit]
+
+    def find_by_library_and_file_hash(
+        self,
+        library_id: str,
+        file_hash: str,
+    ) -> Asset | None:
+        """按素材库 + 文件哈希查找已有素材（去重检测）。"""
+        if not file_hash:
+            return None
+        for asset in self._assets.values():
+            if asset.library_id == library_id and asset.file_hash == file_hash:
+                return asset
+        return None

packages/adapters/in_memory/tag_repository.py

@@ -0,0 +1,40 @@
+"""标签 InMemory 仓储实现。"""
+
+from packages.domain import Tag
+
+
+class InMemoryTagRepository:
+    def __init__(self):
+        self._tags: dict[str, Tag] = {}
+
+    def create(self, tag: Tag) -> Tag:
+        self._tags[tag.id] = tag
+        return tag
+
+    def get(self, tag_id: str) -> Tag | None:
+        return self._tags.get(tag_id)
+
+    def find_by_name(self, user_id: str, name: str) -> Tag | None:
+        for tag in self._tags.values():
+            if tag.user_id == user_id and tag.name == name:
+                return tag
+        return None
+
+    def list_by_user(
+        self,
+        user_id: str,
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Tag]:
+        tags = [tag for tag in self._tags.values() if tag.user_id == user_id]
+        tags.sort(key=lambda t: t.created_at, reverse=True)
+        return tags[skip : skip + limit]
+
+    def count_by_user(self, user_id: str) -> int:
+        return sum(1 for tag in self._tags.values() if tag.user_id == user_id)
+
+    def delete(self, tag_id: str) -> bool:
+        if tag_id in self._tags:
+            del self._tags[tag_id]
+            return True
+        return False

packages/adapters/sqlalchemy_impl/asset_repository.py

@@ -3,7 +3,7 @@ from datetime import datetime, timezone
 
 from sqlalchemy.orm import Session
 
-from packages.adapters.sqlalchemy_impl.models import AssetModel
+from packages.adapters.sqlalchemy_impl.models import AssetModel, AssetTagModel
 from packages.domain import Asset, AssetStatus, ClassificationStatus
 
 
@@ -37,6 +37,22 @@ class SQLAlchemyAssetRepository:
         )
         return [self._to_domain(model) for model in models]
 
+    def find_by_library_and_file_type(
+        self,
+        library_id: str,
+        file_type: str,
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Asset]:
+        models = (
+            self.session.query(AssetModel)
+            .filter(AssetModel.asset_library_id == library_id, AssetModel.file_type == file_type)
+            .offset(skip)
+            .limit(limit)
+            .all()
+        )
+        return [self._to_domain(model) for model in models]
+
     def find_by_id(self, asset_id: str) -> Asset | None:
         model = self.session.query(AssetModel).filter(AssetModel.id == asset_id).first()
         if model is None:
@@ -67,10 +83,13 @@ class SQLAlchemyAssetRepository:
             classification_result=(json.dumps(asset.metadata) if asset.metadata else None),
             quality_score=asset.quality_score,
             uploaded_by_user_id=asset.uploaded_by_user_id or "system",
+            file_hash=asset.file_hash or None,
             created_at=asset.created_at,
             updated_at=now,
         )
         self.session.add(model)
+        self.session.flush()
+        self._sync_asset_tags(asset.id, asset.tag_ids)
         self.session.commit()
         return asset
 
@@ -92,7 +111,10 @@ class SQLAlchemyAssetRepository:
         model.classification_result = json.dumps(asset.metadata) if asset.metadata else None
         model.quality_score = asset.quality_score
         model.uploaded_by_user_id = asset.uploaded_by_user_id or model.uploaded_by_user_id
+        model.file_hash = asset.file_hash or model.file_hash
         model.updated_at = datetime.now(timezone.utc)
+        self.session.flush()
+        self._sync_asset_tags(asset.id, asset.tag_ids)
         self.session.commit()
         return asset
 
@@ -104,6 +126,14 @@ class SQLAlchemyAssetRepository:
             return True
         return False
 
+    def batch_delete(self, asset_ids: list[str]) -> int:
+        """批量删除素材，返回实际删除数量。"""
+        if not asset_ids:
+            return 0
+        count = self.session.query(AssetModel).filter(AssetModel.id.in_(asset_ids)).delete(synchronize_session=False)
+        self.session.commit()
+        return count
+
     def count_by_project(self, project_id: str) -> int:
         return self.session.query(AssetModel).filter(AssetModel.project_id == project_id).count()
 
@@ -179,6 +209,11 @@ class SQLAlchemyAssetRepository:
                 "audio": "audio/mpeg",
                 "image": "image/jpeg",
             }.get(mime_type, mime_type)
+        # 查询关联的 tag_ids
+        tag_ids = [
+            row.tag_id
+            for row in self.session.query(AssetTagModel.tag_id).filter(AssetTagModel.asset_id == model.id).all()
+        ]
         return Asset(
             id=model.id,
             project_id=model.project_id,
@@ -197,7 +232,61 @@ class SQLAlchemyAssetRepository:
             classification_status=ClassificationStatus(model.classification_status),
             quality_score=model.quality_score,
             uploaded_by_user_id=model.uploaded_by_user_id,
+            file_hash=model.file_hash or "",
             metadata=metadata,
+            tag_ids=tag_ids,
             created_at=model.created_at,
             updated_at=model.updated_at,
         )
+
+    def _sync_asset_tags(self, asset_id: str, tag_ids: list[str]) -> None:
+        """同步素材-标签关联表（全量替换）。"""
+        self.session.query(AssetTagModel).filter(AssetTagModel.asset_id == asset_id).delete(synchronize_session=False)
+        for tag_id in tag_ids:
+            self.session.add(AssetTagModel(asset_id=asset_id, tag_id=tag_id))
+
+    def find_by_tag_ids(
+        self,
+        tag_ids: list[str],
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Asset]:
+        """查找包含所有指定标签的素材。"""
+        if not tag_ids:
+            return []
+        from sqlalchemy import func
+
+        # 找出同时拥有所有指定 tag_id 的 asset_id
+        tag_set = set(tag_ids)
+        asset_ids = (
+            self.session.query(AssetTagModel.asset_id)
+            .filter(AssetTagModel.tag_id.in_(tag_set))
+            .group_by(AssetTagModel.asset_id)
+            .having(func.count(AssetTagModel.tag_id) == len(tag_set))
+            .all()
+        )
+        ids = [row[0] for row in asset_ids]
+        if not ids:
+            return []
+        models = self.session.query(AssetModel).filter(AssetModel.id.in_(ids)).offset(skip).limit(limit).all()
+        return [self._to_domain(m) for m in models]
+
+    def find_by_library_and_file_hash(
+        self,
+        library_id: str,
+        file_hash: str,
+    ) -> Asset | None:
+        """按素材库 + 文件哈希查找已有素材（去重检测）。"""
+        if not file_hash:
+            return None
+        model = (
+            self.session.query(AssetModel)
+            .filter(
+                AssetModel.asset_library_id == library_id,
+                AssetModel.file_hash == file_hash,
+            )
+            .first()
+        )
+        if model is None:
+            return None
+        return self._to_domain(model)

packages/adapters/sqlalchemy_impl/generated_video_repository.py

@@ -78,7 +78,7 @@ class SQLAlchemyGeneratedVideoRepository:
 
     def list_by_project(self, project_id: str) -> list[GeneratedVideo]:
         models = self.session.query(GeneratedVideoModel).filter(GeneratedVideoModel.project_id == project_id).all()
-        return [self.get(model.id) for model in models if self.get(model.id) is not None]
+        return [self._to_domain(model) for model in models]
 
     def list_by_generation_task(self, generation_task_id: str) -> list[GeneratedVideo]:
         models = (
@@ -86,4 +86,40 @@ class SQLAlchemyGeneratedVideoRepository:
             .filter(GeneratedVideoModel.generation_task_id == generation_task_id)
             .all()
         )
-        return [self.get(model.id) for model in models if self.get(model.id) is not None]
+        return [self._to_domain(model) for model in models]
+
+    def list_by_batch(self, batch_id: str) -> list[GeneratedVideo]:
+        """通过 batch_id 查找同批次生成的所有视频（跨 generation_task 关联查询）。"""
+        from packages.adapters.sqlalchemy_impl.models import GenerationTaskModel
+
+        task_ids = (
+            self.session.query(GenerationTaskModel.id).filter(GenerationTaskModel.batch_id == batch_id).subquery()
+        )
+        models = (
+            self.session.query(GeneratedVideoModel).filter(GeneratedVideoModel.generation_task_id.in_(task_ids)).all()
+        )
+        return [self._to_domain(model) for model in models]
+
+    @staticmethod
+    def _to_domain(model: GeneratedVideoModel) -> GeneratedVideo:
+        return GeneratedVideo(
+            id=model.id,
+            project_id=model.project_id,
+            generation_task_id=model.generation_task_id,
+            name=model.name,
+            file_url=model.file_url,
+            file_size=int(model.file_size or 0),
+            duration=model.duration,
+            thumbnail_url=model.thumbnail_url,
+            width=int(model.width or 0),
+            height=int(model.height or 0),
+            fps=model.fps,
+            status=getattr(model, "status", "completed"),
+            review_status=getattr(model, "review_status", "pending_review"),
+            generation_params=json.loads(getattr(model, "generation_params", "{}") or "{}"),
+            video_fingerprint=json.loads(getattr(model, "video_fingerprint", "null") or "null"),
+            is_duplicate=getattr(model, "is_duplicate", False),
+            duplicate_of=getattr(model, "duplicate_of", None),
+            generated_at=model.generated_at,
+            created_at=model.created_at,
+        )

packages/adapters/sqlalchemy_impl/generation_task_repository.py

@@ -25,6 +25,8 @@ def _to_domain(model: GenerationTaskModel) -> GenerationTask:
         completed_at=model.completed_at,
         created_by_user_id=model.created_by_user_id,
         source_edit_plan_id=model.source_edit_plan_id or "",
+        asset_select_mode=model.asset_select_mode or "",
+        batch_id=model.batch_id or "",
         created_at=model.created_at,
     )
 
@@ -52,6 +54,8 @@ class SQLAlchemyGenerationTaskRepository:
             completed_at=task.completed_at,
             created_by_user_id=task.created_by_user_id,
             source_edit_plan_id=task.source_edit_plan_id or None,
+            asset_select_mode=task.asset_select_mode or "",
+            batch_id=task.batch_id or "",
             created_at=task.created_at,
         )
         self.session.add(model)
@@ -123,5 +127,7 @@ class SQLAlchemyGenerationTaskRepository:
         model.started_at = task.started_at
         model.completed_at = task.completed_at
         model.source_edit_plan_id = task.source_edit_plan_id or None
+        model.asset_select_mode = task.asset_select_mode or ""
+        model.batch_id = task.batch_id or ""
         self.session.commit()
         return task

packages/adapters/sqlalchemy_impl/ingest_job_repository.py

@@ -17,6 +17,7 @@ class SQLAlchemyIngestJobRepository:
             status=job.status.value,
             error_message=job.error_message,
             result_asset_id=job.result_asset_id,
+            file_hash=job.file_hash,
             created_at=job.created_at,
             updated_at=job.updated_at,
         )
@@ -36,6 +37,7 @@ class SQLAlchemyIngestJobRepository:
             status=IngestJobStatus(model.status),
             error_message=model.error_message,
             result_asset_id=model.result_asset_id,
+            file_hash=model.file_hash or "",
             created_at=model.created_at,
             updated_at=model.updated_at,
         )
@@ -51,6 +53,7 @@ class SQLAlchemyIngestJobRepository:
         model.status = job.status.value
         model.error_message = job.error_message
         model.result_asset_id = job.result_asset_id
+        model.file_hash = job.file_hash
         model.updated_at = job.updated_at
         self.session.commit()
         return job

packages/adapters/sqlalchemy_impl/models.py

@@ -85,11 +85,35 @@ class AssetModel(Base):
     classification_result = Column(Text, nullable=True)
     quality_score = Column(Float, nullable=True)
     uploaded_by_user_id = Column(String(36), nullable=False)
+    file_hash = Column(String(64), nullable=True, index=True)
     extra_meta = Column("metadata", JSON, nullable=False, default=dict)
     created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc), index=True)
     updated_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
 
 
+class TagModel(Base):
+    """标签 ORM 模型。"""
+
+    __tablename__ = "tags"
+
+    id = Column(String(36), primary_key=True)
+    user_id = Column(String(36), nullable=False, index=True)
+    name = Column(String(100), nullable=False)
+    created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
+
+    __table_args__ = (UniqueConstraint("user_id", "name", name="uq_tags_user_name"),)
+
+
+class AssetTagModel(Base):
+    """素材-标签关联表 ORM 模型。"""
+
+    __tablename__ = "asset_tags"
+
+    asset_id = Column(String(36), primary_key=True)
+    tag_id = Column(String(36), primary_key=True)
+    created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
+
+
 class EditTemplateModel(Base):
     """Phase 8 剪辑模板 ORM 模型
 
@@ -187,6 +211,7 @@ class IngestJobModel(Base):
     status = Column(String(20), nullable=False, default="pending")
     error_message = Column(Text, nullable=False, default="")
     result_asset_id = Column(String(32), nullable=False, default="")
+    file_hash = Column(String(64), nullable=True, index=True)
     created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
     updated_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
 
@@ -228,6 +253,8 @@ class GenerationTaskModel(Base):
     completed_at = Column(DateTime, nullable=True)
     created_by_user_id = Column(String(32), nullable=False, default="", index=True)
     source_edit_plan_id = Column(String(32), nullable=True, index=True)
+    asset_select_mode = Column(String(20), nullable=False, default="")
+    batch_id = Column(String(32), nullable=False, default="", index=True)
     extra_meta = Column("metadata", JSON, nullable=False, default=dict)
     created_at = Column(DateTime, nullable=False, default=lambda: datetime.now(timezone.utc))
 

packages/adapters/sqlalchemy_impl/tag_repository.py

@@ -0,0 +1,73 @@
+"""标签 SQLAlchemy 仓储实现。"""
+
+from sqlalchemy.orm import Session
+
+from packages.adapters.sqlalchemy_impl.models import AssetTagModel, TagModel
+from packages.domain import Tag
+
+
+class SQLAlchemyTagRepository:
+    def __init__(self, session: Session):
+        self.session = session
+
+    def create(self, tag: Tag) -> Tag:
+        model = TagModel(
+            id=tag.id,
+            user_id=tag.user_id,
+            name=tag.name,
+            created_at=tag.created_at,
+        )
+        self.session.add(model)
+        self.session.commit()
+        return tag
+
+    def get(self, tag_id: str) -> Tag | None:
+        model = self.session.query(TagModel).filter(TagModel.id == tag_id).first()
+        if model is None:
+            return None
+        return self._to_domain(model)
+
+    def find_by_name(self, user_id: str, name: str) -> Tag | None:
+        model = self.session.query(TagModel).filter(TagModel.user_id == user_id, TagModel.name == name).first()
+        if model is None:
+            return None
+        return self._to_domain(model)
+
+    def list_by_user(
+        self,
+        user_id: str,
+        skip: int = 0,
+        limit: int = 100,
+    ) -> list[Tag]:
+        models = (
+            self.session.query(TagModel)
+            .filter(TagModel.user_id == user_id)
+            .order_by(TagModel.created_at.desc())
+            .offset(skip)
+            .limit(limit)
+            .all()
+        )
+        return [self._to_domain(m) for m in models]
+
+    def count_by_user(self, user_id: str) -> int:
+        return self.session.query(TagModel).filter(TagModel.user_id == user_id).count()
+
+    def delete(self, tag_id: str) -> bool:
+        # 先清理关联表
+        self.session.query(AssetTagModel).filter(AssetTagModel.tag_id == tag_id).delete(synchronize_session=False)
+        model = self.session.query(TagModel).filter(TagModel.id == tag_id).first()
+        if model is None:
+            self.session.commit()
+            return False
+        self.session.delete(model)
+        self.session.commit()
+        return True
+
+    @staticmethod
+    def _to_domain(model: TagModel) -> Tag:
+        return Tag(
+            id=model.id,
+            user_id=model.user_id,
+            name=model.name,
+            created_at=model.created_at,
+        )

packages/adapters/sqlalchemy_impl/user_repository.py

@@ -33,6 +33,8 @@ class SQLAlchemyUserRepository(UserRepository):
         model.max_projects = user.max_projects
         model.max_storage_gb = user.max_storage_gb
         model.is_admin = user.is_admin
+        model.wechat_openid = user.wechat_openid
+        model.wechat_unionid = user.wechat_unionid
         model.created_at = user.created_at
 
         self.session.commit()
@@ -49,6 +51,16 @@ class SQLAlchemyUserRepository(UserRepository):
         model = self.session.query(UserModel).filter(UserModel.username == username.strip()).first()
         return self._to_entity(model)
 
+    def find_by_wechat_openid(self, openid: str) -> User | None:
+        model = self.session.query(UserModel).filter(UserModel.wechat_openid == openid.strip()).first()
+        return self._to_entity(model)
+
+    def find_by_wechat_unionid(self, unionid: str) -> User | None:
+        if not unionid or not unionid.strip():
+            return None
+        model = self.session.query(UserModel).filter(UserModel.wechat_unionid == unionid.strip()).first()
+        return self._to_entity(model)
+
     def find_by_verification_token(self, token: str) -> User | None:
         model = self.session.query(UserModel).filter(UserModel.email_verification_token == token).first()
         return self._to_entity(model)
@@ -87,5 +99,7 @@ class SQLAlchemyUserRepository(UserRepository):
             max_projects=model.max_projects or 3,
             max_storage_gb=model.max_storage_gb or 10,
             is_admin=model.is_admin or False,
+            wechat_openid=model.wechat_openid,
+            wechat_unionid=model.wechat_unionid,
             created_at=model.created_at,
         )

packages/application/auth/wechat_sync_use_case.py

@@ -0,0 +1,206 @@
+"""
+微信同步登录/注册 Use Case
+
+供 BFF 层调用的系统级接口：
+- 根据 openid 查找用户，找到则登录返回 token
+- 没找到则创建新用户并返回 token
+- 支持 unionid 跨应用关联
+"""
+
+import secrets
+from datetime import datetime, timedelta, timezone
+from typing import Optional
+from uuid import uuid4
+
+import jwt as pyjwt
+
+from packages.adapters.redis import get_session_store
+from packages.application.auth.jwt_service import jwt_service
+from packages.domain.entities import User
+
+
+class WechatSyncRequest:
+    """微信同步登录请求"""
+
+    def __init__(
+        self,
+        openid: str,
+        unionid: str = "",
+        nickname: str = "",
+        avatar_url: str = "",
+        source: str = "miniapp",
+    ):
+        self.openid = openid.strip()
+        self.unionid = unionid.strip() if unionid else ""
+        self.nickname = nickname or "微信用户"
+        self.avatar_url = avatar_url or ""
+        self.source = source
+
+
+class WechatSyncResponse:
+    """微信同步登录响应"""
+
+    def __init__(
+        self,
+        access_token: str,
+        refresh_token: str,
+        user_id: str,
+        nickname: str,
+        avatar_url: str,
+        is_new_user: bool,
+        expires_in: int,
+    ):
+        self.access_token = access_token
+        self.refresh_token = refresh_token
+        self.user_id = user_id
+        self.nickname = nickname
+        self.avatar_url = avatar_url
+        self.is_new_user = is_new_user
+        self.expires_in = expires_in
+
+    def to_dict(self) -> dict:
+        return {
+            "access_token": self.access_token,
+            "token": self.access_token,  # 兼容 BFF 层用 token 字段读取
+            "refresh_token": self.refresh_token,
+            "user_id": self.user_id,
+            "user": {
+                "id": self.user_id,
+                "nickname": self.nickname,
+                "avatar_url": self.avatar_url,
+                "display_name": self.nickname,
+            },
+            "user_info": {
+                "id": self.user_id,
+                "nickname": self.nickname,
+                "avatar_url": self.avatar_url,
+                "display_name": self.nickname,
+            },
+            "is_new_user": self.is_new_user,
+            "expires_in": self.expires_in,
+        }
+
+
+class WechatSyncUseCase:
+    """微信同步登录/注册用例
+
+    系统级接口，由 BFF 通过 API Key 调用。
+    职责：根据 openid 查找或创建用户，返回 SaaS token。
+    """
+
+    def __init__(self, user_repository, session_store=None, jwt_secret_key: str | None = None):
+        self.user_repository = user_repository
+        self.session_store = session_store or get_session_store()
+        self.jwt_secret_key = jwt_secret_key or jwt_service.config.SECRET_KEY
+
+    def execute(self, request: WechatSyncRequest) -> tuple[Optional[WechatSyncResponse], Optional[str]]:
+        """
+        执行微信同步登录/注册
+
+        Returns:
+            (响应对象, 错误信息) - 成功则错误信息为 None
+        """
+        try:
+            if not request.openid:
+                return None, "openid is required"
+
+            is_new_user = False
+
+            # 1. 按 openid 查找用户
+            user = self.user_repository.find_by_wechat_openid(request.openid)
+
+            # 2. 如果 openid 没找到，尝试 unionid
+            if not user and request.unionid:
+                user = self.user_repository.find_by_wechat_unionid(request.unionid)
+                if user:
+                    # 找到用户但 openid 为空，绑定一下当前 openid
+                    user.wechat_openid = request.openid
+                    self.user_repository.save(user)
+
+            # 3. 都没找到则创建新用户
+            if not user:
+                user = self._create_wechat_user(request)
+                is_new_user = True
+
+            # 4. 创建 session 并生成 token
+            session_id = secrets.token_urlsafe(16)
+            refresh_token = secrets.token_urlsafe(32)
+
+            now = datetime.now(timezone.utc)
+            access_token_payload = {
+                "sub": user.id,
+                "sid": session_id,
+                "type": "user_auth",
+                "iat": now,
+                "exp": now + timedelta(minutes=jwt_service.config.ACCESS_TOKEN_EXPIRE_MINUTES),
+            }
+            access_token = pyjwt.encode(
+                access_token_payload,
+                self.jwt_secret_key,
+                algorithm=jwt_service.config.ALGORITHM,
+            )
+
+            # 保存 session
+            self.session_store.save_session(
+                session_id=session_id,
+                user_id=user.id,
+                refresh_token=refresh_token,
+                device_info=f"wechat_{request.source}",
+                ip_address="bff_gateway",
+                expires_in_seconds=30 * 24 * 3600,  # 30 天
+            )
+
+            # 更新最后登录信息
+            user.last_login_at = now
+            user.last_login_ip = "bff_gateway"
+            self.user_repository.save(user)
+
+            response = WechatSyncResponse(
+                access_token=access_token,
+                refresh_token=refresh_token,
+                user_id=user.id,
+                nickname=user.display_name,
+                avatar_url="",  # SaaS 用户模型暂存头像，后续可扩展
+                is_new_user=is_new_user,
+                expires_in=jwt_service.config.ACCESS_TOKEN_EXPIRE_MINUTES * 60,
+            )
+
+            return response, None
+
+        except Exception as e:
+            return None, f"Internal error: {str(e)}"
+
+    def _create_wechat_user(self, request: WechatSyncRequest) -> User:
+        """创建微信用户"""
+        user_id = uuid4().hex
+
+        # 生成唯一名和邮箱（微信用户无真实邮箱，用 openid 生成占位）
+        safe_openid = request.openid.replace("-", "_")[:20]
+        username = f"wx_{safe_openid}"
+        email = f"{safe_openid}@wechat.local"
+
+        # 确保 username 唯一
+        suffix = 0
+        while self.user_repository.find_by_username(username):
+            suffix += 1
+            username = f"wx_{safe_openid}_{suffix}"
+
+        # 随机密码（微信用户不用密码登录）
+        random_password = secrets.token_urlsafe(32)
+        from packages.application.auth.password_hasher import password_hasher
+
+        password_hash = password_hasher.hash_password(random_password)
+
+        user = User(
+            id=user_id,
+            email=email,
+            username=username,
+            display_name=request.nickname or "微信用户",
+            password_hash=password_hash,
+            email_verified=True,  # 微信登录视为已验证
+            wechat_openid=request.openid,
+            wechat_unionid=request.unionid or None,
+        )
+
+        self.user_repository.save(user)
+        return user

packages/application/generation_tasks.py

@@ -19,6 +19,8 @@ class CreateGenerationTaskCommand:
     voice_ids: list[str] = field(default_factory=list)
     created_by_user_id: str = ""
     source_edit_plan_id: str = ""
+    asset_select_mode: str = ""
+    batch_id: str = ""
 
 
 class CreateGenerationTaskUseCase:
@@ -44,6 +46,8 @@ class CreateGenerationTaskUseCase:
             completed_at=None,
             created_by_user_id=command.created_by_user_id,
             source_edit_plan_id=command.source_edit_plan_id,
+            asset_select_mode=command.asset_select_mode,
+            batch_id=command.batch_id,
         )
         return self.generation_task_repository.create(task)
 

packages/application/ingest_jobs.py

@@ -11,6 +11,7 @@ class SubmitIngestJobCommand:
     project_id: str
     library_id: str
     storage_key: str
+    file_hash: str = ""
 
 
 class SubmitIngestJobUseCase:
@@ -22,5 +23,6 @@ class SubmitIngestJobUseCase:
             project_id=command.project_id,
             library_id=command.library_id,
             storage_key=command.storage_key,
+            file_hash=command.file_hash,
         )
         return self.ingest_job_repository.create(job)

packages/application/tts_job/audio_merger.py

@@ -0,0 +1,96 @@
+"""FFmpeg 音频合并器 — P1 长文本分段合成。
+
+将多个分段音频文件合并为一个完整音频文件。
+"""
+
+from __future__ import annotations
+
+import logging
+import os
+import shutil
+import subprocess
+import tempfile
+
+logger = logging.getLogger(__name__)
+
+
+class AudioMergeError(Exception):
+    """音频合并异常。"""
+
+    pass
+
+
+class AudioMerger:
+    """使用 FFmpeg 合并多个音频文件。"""
+
+    def merge(self, audio_paths: list[str], output_format: str = "mp3") -> bytes:
+        """合并多个音频文件，返回合并后的音频数据。
+
+        使用 FFmpeg concat demuxer 按顺序拼接音频。
+        所有输入文件必须为相同格式和采样率。
+
+        Args:
+            audio_paths: 音频文件路径列表（按合成顺序）
+            output_format: 输出格式（mp3/wav/pcm）
+
+        Returns:
+            合并后的音频文件字节数据
+
+        Raises:
+            AudioMergeError: 合并失败
+        """
+        if not audio_paths:
+            raise AudioMergeError("没有可合并的音频文件")
+
+        if len(audio_paths) == 1:
+            with open(audio_paths[0], "rb") as f:
+                return f.read()
+
+        temp_dir = tempfile.mkdtemp(prefix="tts_merge_")
+        try:
+            # 生成 concat demuxer 列表文件
+            list_path = os.path.join(temp_dir, "concat_list.txt")
+            with open(list_path, "w") as f:
+                for path in audio_paths:
+                    # FFmpeg concat 文件需要 file: 前缀，路径中的 ' 和 \n 需转义
+                    escaped = path.replace("'", "'\\''").replace("\n", "\\n")
+                    f.write(f"file '{escaped}'\n")
+
+            output_path = os.path.join(temp_dir, f"merged.{output_format}")
+
+            cmd = [
+                "ffmpeg",
+                "-y",
+                "-f",
+                "concat",
+                "-safe",
+                "0",
+                "-i",
+                list_path,
+                "-c",
+                "copy",
+                output_path,
+            ]
+
+            result = subprocess.run(
+                cmd,
+                capture_output=True,
+                text=True,
+                timeout=120,
+            )
+
+            if result.returncode != 0:
+                logger.error(f"FFmpeg 合并失败: stderr={result.stderr}")
+                raise AudioMergeError(f"FFmpeg 合并失败: {result.stderr[:500]}")
+
+            with open(output_path, "rb") as f:
+                return f.read()
+
+        except subprocess.TimeoutExpired:
+            raise AudioMergeError("FFmpeg 合并超时（120 秒）")
+        except AudioMergeError:
+            raise
+        except Exception as e:
+            raise AudioMergeError(f"音频合并失败: {e}")
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)

packages/application/tts_job/streaming_service.py

@@ -0,0 +1,251 @@
+"""P2: TTS 流式合成服务 — WebSocket 实时音频推送。
+
+通过 WebSocket 将合成音频以二进制帧实时推送给客户端。
+- 短文本（≤500 字）：合成完整音频后分块推送
+- 长文本（>500 字）：分段并发合成，逐段推送音频
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+from typing import Any, Optional
+
+import httpx
+
+from packages.application.cosyvoice_service import CosyVoiceError, CosyVoiceService
+from packages.application.tts_job.text_splitter import split_text
+
+logger = logging.getLogger(__name__)
+
+# WebSocket 二进制帧块大小（4KB）
+_AUDIO_CHUNK_SIZE = 4096
+# 分段并发上限
+_MAX_STREAMING_SEGMENT_WORKERS = 5
+# 长文本分段阈值
+_SEGMENT_THRESHOLD = 500
+# WebSocket 最大文本长度
+_MAX_TEXT_LENGTH = 10000
+
+
+class TTSStreamingError(Exception):
+    """TTS 流式合成异常。"""
+
+    pass
+
+
+class TTSStreamingService:
+    """TTS 流式合成服务。
+
+    通过 WebSocket 实时推送合成音频。
+    使用 CosyVoiceService（同步 REST API）合成，
+    通过 asyncio.to_thread 桥接到异步 WebSocket。
+    """
+
+    def __init__(self, cosyvoice_service: CosyVoiceService) -> None:
+        self._cosyvoice = cosyvoice_service
+
+    async def synthesize_and_stream(self, websocket: Any, params: dict) -> None:
+        """根据文本长度选择流式合成策略。
+
+        Args:
+            websocket: FastAPI WebSocket 连接
+            params: 合成参数（text, voice_id, sample_rate, format, speed）
+        """
+        text = params.get("text", "")
+        if not text:
+            await self._send_json(websocket, {"type": "error", "message": "文本不能为空"})
+            return
+
+        if len(text) > _MAX_TEXT_LENGTH:
+            await self._send_json(
+                websocket,
+                {"type": "error", "message": f"文本过长，最大 {_MAX_TEXT_LENGTH} 字"},
+            )
+            return
+
+        if len(text) <= _SEGMENT_THRESHOLD:
+            await self._stream_short_text(websocket, params)
+        else:
+            await self._stream_long_text(websocket, params)
+
+    # ── 短文本流式合成 ────────────────────────────────────────
+
+    async def _stream_short_text(self, websocket: Any, params: dict) -> None:
+        """短文本：合成完整音频后分块推送。"""
+        text = params["text"]
+        voice_id = params.get("voice_id", "")
+        sample_rate = params.get("sample_rate", 0)
+        audio_format = params.get("format", "mp3")
+        speed = params.get("speed", 1.0)
+
+        await self._send_json(
+            websocket,
+            {"type": "started", "segment_count": 1, "total_segments": 1},
+        )
+
+        # 在线程池中执行同步合成
+        try:
+            result = await asyncio.to_thread(
+                self._cosyvoice.submit_synthesize_task,
+                text=text,
+                voice_id=voice_id,
+                sample_rate=sample_rate,
+                format=audio_format,
+                speed=speed,
+            )
+        except CosyVoiceError as e:
+            logger.error(f"流式合成失败: {e}")
+            await self._send_json(websocket, {"type": "error", "message": str(e)})
+            return
+        except Exception as e:
+            logger.error(f"流式合成意外错误: {e}")
+            await self._send_json(websocket, {"type": "error", "message": f"合成失败: {e}"})
+            return
+
+        audio_url = result.get("audio_url", "")
+        if not audio_url:
+            await self._send_json(websocket, {"type": "error", "message": "合成未返回音频 URL"})
+            return
+
+        # 下载并流式推送音频
+        try:
+            audio_data = await asyncio.to_thread(self._download_audio, audio_url)
+            total_bytes = await self._stream_audio_chunks(websocket, audio_data)
+
+            await self._send_json(
+                websocket,
+                {
+                    "type": "done",
+                    "duration": result.get("duration", 0.0),
+                    "file_size": total_bytes,
+                    "format": audio_format,
+                },
+            )
+        except Exception as e:
+            logger.error(f"音频流式推送失败: {e}")
+            await self._send_json(websocket, {"type": "error", "message": f"音频推送失败: {e}"})
+
+    # ── 长文本分段流式合成 ────────────────────────────────────
+
+    async def _stream_long_text(self, websocket: Any, params: dict) -> None:
+        """长文本：分段并发合成，逐段推送音频。"""
+        text = params["text"]
+        voice_id = params.get("voice_id", "")
+        sample_rate = params.get("sample_rate", 0)
+        audio_format = params.get("format", "mp3")
+        speed = params.get("speed", 1.0)
+
+        segments = split_text(text, max_chars=_SEGMENT_THRESHOLD)
+        segment_count = len(segments)
+
+        logger.info(f"流式分段合成: 原文={len(text)}字, 段数={segment_count}")
+
+        await self._send_json(
+            websocket,
+            {"type": "started", "segment_count": segment_count, "total_segments": segment_count},
+        )
+
+        # 并发合成所有分段，按顺序流式推送
+        queue: asyncio.Queue[tuple[int, Optional[bytes], Optional[str]]] = asyncio.Queue()
+        completed_count = 0
+
+        async def _synthesize_one(idx: int, seg_text: str) -> None:
+            """合成单个分段并放入队列。"""
+            try:
+                result = await asyncio.to_thread(
+                    self._cosyvoice.submit_synthesize_task,
+                    text=seg_text,
+                    voice_id=voice_id,
+                    sample_rate=sample_rate,
+                    format=audio_format,
+                    speed=speed,
+                )
+                audio_url = result.get("audio_url", "")
+                if audio_url:
+                    audio_data = await asyncio.to_thread(self._download_audio, audio_url)
+                    await queue.put((idx, audio_data, None))
+                else:
+                    await queue.put((idx, None, "合成未返回音频 URL"))
+            except Exception as e:
+                await queue.put((idx, None, str(e)))
+
+        # 启动并发合成任务
+        workers = [asyncio.create_task(_synthesize_one(idx, seg)) for idx, seg in enumerate(segments)]
+
+        # 按顺序消费队列，流式推送
+        total_bytes = 0
+        total_duration = 0.0
+        consumed = 0
+
+        try:
+            while consumed < segment_count:
+                idx, audio_data, error = await queue.get()
+                consumed += 1
+
+                if error:
+                    logger.error(f"分段 {idx + 1} 合成失败: {error}")
+                    await self._send_json(
+                        websocket,
+                        {"type": "error", "message": f"分段 {idx + 1} 合成失败: {error}"},
+                    )
+                    # 取消剩余 worker
+                    for w in workers:
+                        w.cancel()
+                    return
+
+                if audio_data:
+                    seg_bytes = await self._stream_audio_chunks(websocket, audio_data)
+                    total_bytes += seg_bytes
+
+                await self._send_json(
+                    websocket,
+                    {"type": "segment_done", "segment": idx + 1, "total": segment_count},
+                )
+
+            # 等待所有 worker 完成
+            await asyncio.gather(*workers, return_exceptions=True)
+
+            await self._send_json(
+                websocket,
+                {
+                    "type": "done",
+                    "duration": total_duration,
+                    "file_size": total_bytes,
+                    "format": audio_format,
+                },
+            )
+
+        except Exception as e:
+            logger.error(f"流式分段推送失败: {e}")
+            await self._send_json(websocket, {"type": "error", "message": f"推送失败: {e}"})
+            for w in workers:
+                w.cancel()
+
+    # ── 工具方法 ────────────────────────────────────────────
+
+    def _download_audio(self, url: str) -> bytes:
+        """下载音频数据。"""
+        resp = httpx.get(url, timeout=60.0, follow_redirects=True)
+        resp.raise_for_status()
+        return resp.content
+
+    async def _stream_audio_chunks(self, websocket: Any, audio_data: bytes) -> int:
+        """将音频数据分块通过 WebSocket 推送。
+
+        Returns:
+            推送的总字节数
+        """
+        total = 0
+        for offset in range(0, len(audio_data), _AUDIO_CHUNK_SIZE):
+            chunk = audio_data[offset : offset + _AUDIO_CHUNK_SIZE]
+            await websocket.send_bytes(chunk)
+            total += len(chunk)
+        return total
+
+    async def _send_json(self, websocket: Any, data: dict) -> None:
+        """安全发送 JSON 帧。"""
+        try:
+            await websocket.send_json(data)
+        except Exception:
+            pass

packages/application/tts_job/text_splitter.py

@@ -0,0 +1,70 @@
+"""长文本分段工具 — P1 长文本分段合成。
+
+将超过阈值的文本按句子边界分段，供 CosyVoice 并发合成后合并。
+"""
+
+from __future__ import annotations
+
+# 中文句子结束符（含全角/半角）
+_SENTENCE_ENDS = frozenset("。！？；\n.!?;")
+
+
+def split_text(text: str, max_chars: int = 500) -> list[str]:
+    """将文本分段，每段不超过 max_chars 个字符。
+
+    优先在句子边界（句号、问号、感叹号、换行符）处分段。
+    若单个句子超过 max_chars，则在逗号等次级标点处拆分。
+    若仍超长，则硬切。
+
+    Args:
+        text: 待分段文本
+        max_chars: 每段最大字符数
+
+    Returns:
+        分段列表，每段 ≤ max_chars。文本为空时返回空列表。
+    """
+    text = text.strip()
+    if not text:
+        return []
+    if len(text) <= max_chars:
+        return [text]
+
+    segments: list[str] = []
+    current = ""
+
+    for char in text:
+        current += char
+        if char in _SENTENCE_ENDS and len(current) >= 50:
+            # 句子边界且长度合理，切段
+            segments.append(current.strip())
+            current = ""
+        elif len(current) >= max_chars:
+            # 达到上限，强制切段
+            segments.append(current.strip())
+            current = ""
+
+    if current.strip():
+        segments.append(current.strip())
+
+    # 合并过短的段（< 50 字符且不是最后一段），减少 API 调用次数
+    merged: list[str] = []
+    buffer = ""
+    for seg in segments:
+        if buffer:
+            combined = buffer + seg
+            if len(combined) <= max_chars:
+                buffer = combined
+                continue
+            merged.append(buffer)
+            buffer = ""
+        if len(seg) < 50:
+            buffer = seg
+        else:
+            merged.append(seg)
+    if buffer:
+        if merged and len(merged[-1]) + len(buffer) <= max_chars:
+            merged[-1] = merged[-1] + buffer
+        else:
+            merged.append(buffer)
+
+    return [s for s in merged if s]

packages/application/tts_job/workflow.py

@@ -9,19 +9,35 @@
 
 from __future__ import annotations
 
+import io
 import logging
+import os
+import shutil
+import tempfile
+import time
+from concurrent.futures import ThreadPoolExecutor, as_completed
 from typing import Optional
 
+import httpx
+
 from packages.application.cosyvoice_service import (
     CosyVoiceAuthError,
     CosyVoiceError,
     CosyVoiceService,
 )
+from packages.application.tts_job.audio_merger import AudioMergeError, AudioMerger
+from packages.application.tts_job.text_splitter import split_text
 from packages.domain.tts_job import TTSJob, TTSJobStatus
 from packages.ports.tts_job_repository import TTSJobRepository
+from packages.shared.storage import SharedStorageService, get_shared_storage_service
 
 logger = logging.getLogger(__name__)
 
+# 长文本分段阈值：超过此字符数自动分段合成
+_SEGMENT_THRESHOLD = 500
+# 分段并发上限
+_MAX_SEGMENT_WORKERS = 5
+
 
 class TTSWorkflowError(Exception):
     """TTS 合成工作流异常。"""
@@ -46,9 +62,55 @@ class TTSWorkflowService:
         self,
         repository: TTSJobRepository,
         cosyvoice_service: CosyVoiceService,
+        storage_service: Optional[SharedStorageService] = None,
     ) -> None:
         self.repository = repository
         self.cosyvoice_service = cosyvoice_service
+        self._storage_service = storage_service
+
+    @property
+    def _storage(self) -> SharedStorageService:
+        if self._storage_service is None:
+            self._storage_service = get_shared_storage_service()
+        return self._storage_service
+
+    def _transfer_audio_to_oss(
+        self,
+        temp_url: str,
+        user_id: str,
+        job_id: str,
+        audio_format: str = "mp3",
+    ) -> tuple[str, str]:
+        """下载 CosyVoice 临时音频并转存到 OSS。
+
+        Returns:
+            (permanent_url, storage_key) 元组。
+            转存失败时回退到原始临时 URL，storage_key 为空字符串。
+        """
+        storage_key = f"tts-outputs/{user_id}/{job_id}.{audio_format}"
+        content_type_map = {
+            "mp3": "audio/mpeg",
+            "wav": "audio/wav",
+            "pcm": "audio/pcm",
+            "opus": "audio/opus",
+        }
+        content_type = content_type_map.get(audio_format, "application/octet-stream")
+
+        try:
+            # 下载临时音频
+            resp = httpx.get(temp_url, timeout=60.0, follow_redirects=True)
+            resp.raise_for_status()
+            audio_data = resp.content
+
+            # 上传到 OSS
+            file_obj = io.BytesIO(audio_data)
+            permanent_url = self._storage.upload_file(file_obj, storage_key, content_type=content_type)
+            logger.info(f"音频转存 OSS 成功: job_id={job_id}, " f"storage_key={storage_key}, size={len(audio_data)}")
+            return permanent_url, storage_key
+
+        except Exception as e:
+            logger.warning(f"音频转存 OSS 失败，使用临时 URL: " f"job_id={job_id}, error={e}")
+            return temp_url, ""
 
     def start_synthesis(
         self,
@@ -80,6 +142,10 @@ class TTSWorkflowService:
         job.mark_processing()
         job = self.repository.update(job)
 
+        # 长文本自动分段合成
+        if len(job.input_text) > _SEGMENT_THRESHOLD:
+            return self._start_segment_synthesis(job)
+
         try:
             submit_result = self.cosyvoice_service.submit_synthesize_task(
                 text=job.input_text,
@@ -93,17 +159,19 @@ class TTSWorkflowService:
             job_metadata["cosyvoice_task_id"] = submit_result.get("task_id", "")
             job_metadata["cosyvoice_request_id"] = submit_result.get("request_id", "")
 
-            # 如果 CosyVoice 同步返回了 audio_url，直接标记完成
+            # 如果 CosyVoice 同步返回了 audio_url，转存 OSS 后标记完成
             audio_url = submit_result.get("audio_url", "")
             if audio_url:
+                permanent_url, storage_key = self._transfer_audio_to_oss(audio_url, job.user_id, job.id, job.format)
                 job.mark_completed(
-                    output_audio_url=audio_url,
+                    output_audio_url=permanent_url,
+                    output_audio_key=storage_key,
                     duration=submit_result.get("duration", 0.0),
                     file_size=submit_result.get("file_size", 0),
                 )
                 job.metadata = job_metadata
                 job = self.repository.update(job)
-                logger.info(f"TTS 合成同步完成: job_id={job.id}, audio_url={audio_url}")
+                logger.info(f"TTS 合成同步完成: job_id={job.id}, audio_url={permanent_url}")
                 return job
 
             job.metadata = job_metadata
@@ -133,6 +201,11 @@ class TTSWorkflowService:
         if job is None:
             raise TTSJobNotFoundError(f"TTS job {job_id} not found")
 
+        # 检查是否为分段合成任务
+        segment_task_ids = (job.metadata or {}).get("segment_task_ids", [])
+        if segment_task_ids:
+            return self._poll_segment_tasks(job)
+
         task_id = (job.metadata or {}).get("cosyvoice_task_id", "")
         if not task_id:
             raise ValueError(f"TTSJob {job_id} has no cosyvoice_task_id in metadata")
@@ -171,13 +244,17 @@ class TTSWorkflowService:
         if job is None:
             raise TTSJobNotFoundError(f"TTS job {job_id} not found")
 
+        # 转存音频到 OSS，获取永久 URL
+        permanent_url, storage_key = self._transfer_audio_to_oss(audio_url, job.user_id, job.id, job.format)
+
         job.mark_completed(
-            output_audio_url=audio_url,
+            output_audio_url=permanent_url,
+            output_audio_key=storage_key,
             duration=duration,
             file_size=file_size,
         )
         job = self.repository.update(job)
-        logger.info(f"TTS 合成成功: job_id={job_id}, audio_url={audio_url}")
+        logger.info(f"TTS 合成成功: job_id={job_id}, audio_url={permanent_url}")
         return job
 
     def process_synthesis_failure(self, job_id: str, error_message: str) -> TTSJob:
@@ -201,3 +278,224 @@ class TTSWorkflowService:
         job = self.repository.update(job)
         logger.error(f"TTS 合成失败: job_id={job_id}, error={error_message}")
         return job
+
+    # ── P1: 长文本分段合成 ─────────────────────────────────────
+
+    def _upload_merged_to_oss(
+        self, merged_data: bytes, user_id: str, job_id: str, audio_format: str
+    ) -> tuple[str, str]:
+        """上传合并后的音频数据到 OSS。
+
+        Returns:
+            (permanent_url, storage_key) 元组。
+            上传失败时返回 ("", "")。
+        """
+        storage_key = f"tts-outputs/{user_id}/{job_id}.{audio_format}"
+        content_type_map = {
+            "mp3": "audio/mpeg",
+            "wav": "audio/wav",
+            "pcm": "audio/pcm",
+            "opus": "audio/opus",
+        }
+        content_type = content_type_map.get(audio_format, "application/octet-stream")
+        try:
+            file_obj = io.BytesIO(merged_data)
+            permanent_url = self._storage.upload_file(file_obj, storage_key, content_type=content_type)
+            return permanent_url, storage_key
+        except Exception as e:
+            logger.warning(f"分段合并音频转存 OSS 失败: job_id={job_id}, error={e}")
+            return "", ""
+
+    def _start_segment_synthesis(self, job: TTSJob) -> TTSJob:
+        """长文本分段合成入口。
+
+        将文本分段后并发提交到 CosyVoice，根据同步/异步结果走不同路径。
+        """
+        segments = split_text(job.input_text, max_chars=_SEGMENT_THRESHOLD)
+        logger.info(f"长文本分段合成: job_id={job.id}, " f"原文={len(job.input_text)}字, 段数={len(segments)}")
+
+        # 记录分段信息到 metadata
+        job_metadata = dict(job.metadata)
+        job_metadata["segment_count"] = len(segments)
+
+        # 并发提交所有分段
+        results = self._submit_segments_concurrent(segments, job)
+        if results is None:
+            # 提交阶段已失败，_submit_segments_concurrent 内部已标记 failed
+            return self.repository.get(job.id)
+
+        # 判断同步还是异步
+        has_audio_urls = any(r.get("audio_url", "") for r in results)
+        has_task_ids = any(r.get("task_id", "") for r in results)
+
+        if has_audio_urls and not has_task_ids:
+            # 所有分段同步返回音频，直接合并
+            return self._process_segments_sync(job, results)
+
+        # 异步路径：保存各分段的 task_id 供后续轮询
+        segment_task_ids = [r.get("task_id", "") for r in results]
+        segment_audio_urls = [r.get("audio_url", "") for r in results]
+        job_metadata["segment_task_ids"] = segment_task_ids
+        job_metadata["segment_audio_urls"] = segment_audio_urls
+        job_metadata["segment_format"] = job.format
+
+        job.metadata = job_metadata
+        job = self.repository.update(job)
+        logger.info(f"分段合成任务已提交（异步）: job_id={job.id}, " f"段数={len(segments)}")
+        return job
+
+    def _submit_segments_concurrent(self, segments: list[str], job: TTSJob) -> list[dict] | None:
+        """并发提交分段合成任务。
+
+        Returns:
+            各分段的结果列表（保持顺序），提交失败时返回 None。
+        """
+        max_workers = min(len(segments), _MAX_SEGMENT_WORKERS)
+        results: list[dict | None] = [None] * len(segments)
+
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            future_to_idx = {}
+            for idx, segment_text in enumerate(segments):
+                future = executor.submit(
+                    self.cosyvoice_service.submit_synthesize_task,
+                    text=segment_text,
+                    voice_id=job.voice_id,
+                    sample_rate=job.sample_rate,
+                    format=job.format,
+                )
+                future_to_idx[future] = idx
+
+            for future in as_completed(future_to_idx):
+                idx = future_to_idx[future]
+                try:
+                    results[idx] = future.result()
+                except Exception as e:
+                    logger.error(f"分段合成提交失败: job_id={job.id}, " f"segment={idx}, error={e}")
+                    self._handle_segment_failure(job, f"分段 {idx + 1} 合成提交失败: {e}")
+                    return None
+
+        return results  # type: ignore[return-value]
+
+    def _process_segments_sync(self, job: TTSJob, results: list[dict]) -> TTSJob:
+        """同步路径：所有分段已返回 audio_url，下载合并后转存 OSS。"""
+        merged_data, total_duration = self._download_and_merge_segments(results, job)
+
+        # 直接上传合并后的音频 bytes 到 OSS
+        permanent_url, storage_key = self._upload_merged_to_oss(merged_data, job.user_id, job.id, job.format)
+
+        job.mark_completed(
+            output_audio_url=permanent_url,
+            output_audio_key=storage_key,
+            duration=total_duration,
+            file_size=len(merged_data),
+        )
+        job = self.repository.update(job)
+        logger.info(f"分段合成完成: job_id={job.id}, " f"merged_size={len(merged_data)}, duration={total_duration:.1f}")
+        return job
+
+    def _download_and_merge_segments(self, results: list[dict], job: TTSJob) -> tuple[bytes, float]:
+        """下载各分段音频并合并。
+
+        Returns:
+            (merged_audio_bytes, total_duration)
+        """
+        temp_dir = tempfile.mkdtemp(prefix="tts_segments_")
+        try:
+            audio_paths: list[str] = []
+            total_duration = 0.0
+
+            for idx, result in enumerate(results):
+                audio_url = result.get("audio_url", "")
+                if not audio_url:
+                    raise TTSWorkflowError(f"分段 {idx + 1} 没有返回 audio_url")
+
+                total_duration += result.get("duration", 0.0)
+
+                # 下载分段音频到临时文件
+                resp = httpx.get(audio_url, timeout=60.0, follow_redirects=True)
+                resp.raise_for_status()
+
+                seg_path = os.path.join(temp_dir, f"seg_{idx:03d}.{job.format}")
+                with open(seg_path, "wb") as f:
+                    f.write(resp.content)
+                audio_paths.append(seg_path)
+
+            # 合并
+            merger = AudioMerger()
+            merged_data = merger.merge(audio_paths, output_format=job.format)
+            return merged_data, total_duration
+
+        finally:
+            shutil.rmtree(temp_dir, ignore_errors=True)
+
+    def _poll_segment_tasks(self, job: TTSJob) -> TTSJob:
+        """轮询所有分段异步任务，全部完成后合并音频。"""
+        segment_task_ids: list[str] = (job.metadata or {}).get("segment_task_ids", [])
+        segment_audio_urls: list[str] = (job.metadata or {}).get("segment_audio_urls", [])
+        segment_count = len(segment_task_ids)
+
+        poll_start = time.monotonic()
+        poll_timeout = 300.0  # 分段任务超时更长
+        poll_interval = 2.0
+
+        while time.monotonic() - poll_start < poll_timeout:
+            all_done = True
+            results: list[dict | None] = [None] * segment_count
+
+            for idx, task_id in enumerate(segment_task_ids):
+                # 已经有音频的分段跳过轮询
+                if idx < len(segment_audio_urls) and segment_audio_urls[idx]:
+                    results[idx] = {
+                        "audio_url": segment_audio_urls[idx],
+                        "duration": 0.0,
+                        "file_size": 0,
+                    }
+                    continue
+
+                try:
+                    result = self.cosyvoice_service.poll_synthesize_task(task_id, timeout=poll_timeout)
+                    results[idx] = result
+                except Exception as e:
+                    logger.error(f"分段任务轮询失败: job_id={job.id}, " f"segment={idx}, error={e}")
+                    self._handle_segment_failure(job, f"分段 {idx + 1} 轮询失败: {e}")
+                    return self.repository.get(job.id)
+
+                if results[idx] is None:
+                    all_done = False
+
+            if all_done and all(r is not None for r in results):
+                # 所有分段完成，下载合并
+                try:
+                    merged_data, total_duration = self._download_and_merge_segments(results, job)
+
+                    # 转存 OSS
+                    permanent_url, storage_key = self._upload_merged_to_oss(
+                        merged_data, job.user_id, job.id, job.format
+                    )
+
+                    job.mark_completed(
+                        output_audio_url=permanent_url,
+                        output_audio_key=storage_key,
+                        duration=total_duration,
+                        file_size=len(merged_data),
+                    )
+                    job = self.repository.update(job)
+                    logger.info(f"分段合成轮询完成: job_id={job.id}, " f"merged_size={len(merged_data)}")
+                    return job
+
+                except Exception as e:
+                    self._handle_segment_failure(job, f"分段合并失败: {e}")
+                    return self.repository.get(job.id)
+
+            # 等待后重试
+            time.sleep(poll_interval)
+
+        # 超时
+        self._handle_segment_failure(job, "分段合成轮询超时（300 秒）")
+        return self.repository.get(job.id)
+
+    def _handle_segment_failure(self, job: TTSJob, error_message: str) -> None:
+        """分段合成失败处理。"""
+        job.mark_failed(error_message)
+        self.repository.update(job)
+        logger.error(f"分段合成失败: job_id={job.id}, error={error_message}")

packages/domain/__init__.py

@@ -24,6 +24,7 @@ from .entities import (
 from .generated_video import GeneratedVideo
 from .generation_task import GenerationTask, GenerationTaskStatus
 from .job import Job, JobStatus, JobType
+from .tag import Tag
 from .template_clip_config import ClipType, TemplateClipConfig, TransitionEffect
 from .title_library import TitleLibraryItem
 from .voice_library import VoiceLibraryItem
@@ -56,6 +57,7 @@ __all__ = [
     "JobStatus",
     "JobType",
     "Project",
+    "Tag",
     "TemplateClipConfig",
     "TransitionEffect",
     "User",