Lov Video Chapter

视频章节师 · Video Chapters

Build a complete chapter-bar workflow from semantic planning to final video. Keep every stage connected through one chapter-project.json.

Route the request

• Read skills/chapter-plan/SKILL.md for subtitle analysis and editorial cuts.
• Read skills/chapter-design/SKILL.md for React Studio editing and styling.
• Read skills/chapter-render/SKILL.md for transparent overlay or burned video.
• Read skills/chapter-export/SKILL.md for 剪映/CapCut and other editor packages.
• Run only the stages the user requests. Run all four for an end-to-end request.

End-to-end workflow

1. Analyze the subtitles

python3 scripts/subtitle_chapters.py \
  --input "/path/to/subtitles.srt" \
  --segments 5 \
  --output "/tmp/video-chapter-analysis.md"

Read the analysis pack completely. Select semantic transitions rather than equal-duration cuts. Start at 00:00, align later cuts to subtitle cues, and avoid cutting a sentence.

2. Create the project

Write a UTF-8 chapter list:

00:00 开场与目标 | 展示成片并说明这次要完成什么
04:16 挑选并改造 Skill | 判断现有工具并完成适配

Create the canonical project:

python3 scripts/chapter_project.py create \
  --chapters "/path/to/chapters.txt" \
  --video "/path/to/video.mp4" \
  --output "/path/to/chapter-project.json"

Read references/project-format.md when editing project JSON directly.

3. Refine in React Studio

cd studio
npm install
npm run dev

Import chapter-project.json, select the local source video, edit chapter boundaries and titles, adjust styling, then export the updated JSON. Treat the Studio preview as the visual approval surface.

4. Render

Install Pillow once when needed:

python3 -m pip install -r requirements.txt

Create a transparent overlay:

python3 scripts/render_chapter_bar.py overlay \
  --project "/path/to/chapter-project.json" \
  --output "/path/to/chapter-overlay.mov"

Burn the overlay into the source video:

python3 scripts/render_chapter_bar.py burn \
  --project "/path/to/chapter-project.json" \
  --output "/path/to/video-with-chapters.mp4"

Create a 剪映-ready package:

python3 scripts/render_chapter_bar.py package \
  --project "/path/to/chapter-project.json" \
  --output "/path/to/chapter-package"

Validation

Run before delivery:

python3 scripts/chapter_project.py validate \
  --project "/path/to/chapter-project.json"

Confirm:

• chapters are chronological and contiguous;
• the final chapter ends at the project duration;
• titles match the following content;
• the Studio preview matches the requested visual direction;
• the overlay contains an alpha channel;
• the burned MP4 preserves source audio;
• the editor package contains JSON, timestamps, CSV, overlay, and instructions.

Output policy

• Prefer a transparent ProRes 4444 MOV for 剪映/CapCut.
• Use burned MP4 when the user wants a final publishable file.
• Describe transparent-video import as asset integration.
• Keep editor-specific draft writers in an experimental adapter layer.
• Preserve explicit paths; never assume a private workspace.

Runtime context (shared)

运行前读取本 Skill 包的 skill.yaml，由宿主提供 skill-runtime/v1 上下文。字段解析顺序为：当前请求、项目上下文、个人 Preferences、品牌 Profile、通用默认值。

• 只使用 Manifest 声明的字段；Profile 保存公开品牌事实，Preferences 保存个人工作偏好。
• required: true 字段缺失时，按 Manifest 的问题配置向用户提出一个聚焦问题；用户明确同意后再保存回答。
• 报错提供可复制的 context_id、字段路径与来源，诊断内容避开秘密、完整私人路径和原始配置。

通用反馈闭环

用户在 Skill 驱动任务中提出修改意见时，继续当前产物前必须执行：

1. 先判断意见是 task-specific（仅本次）还是 reusable（可跨任务复用）。
2. task-specific 只修改当前任务，不改 Skill。
3. reusable 先确定作用域：领域规则先更新对应 canonical Skill；适用于所有 Skill 的规则先更新共享规范。
4. 完成规则更新、版本、lint 与分发核验后，再把修改应用到当前任务。
5. reusable 修改会使此前的“确认”“继续”“发吧”失效；完成当前产物修改和回读后必须停下，等待用户下一步指示，不自动进入发布、提交或其他外部写入。