Unified workflow combining video-clean and video-add-chapters with transcript reuse and chapter remapping
Unified workflow combining video-clean and video-add-chapters with transcript reuse and chapter remapping
npx skills add https://github.com/jykim/claude-obsidian-skills --skill video-full-process复制此命令并粘贴到 Claude Code 中以安装该技能
Interact with Obsidian using the official CLI to read, search, create, and manage notes, daily notes, tasks, properties, and more. Use when the user wants to query or manipulate vault content via the obsidian command.
고비 데스크탑 3.0 대화형 음성 온보딩 가이드
Extract YouTube video transcripts and generate AI-powered summaries in any language. Converts videos to structured markdown documents with summaries, key points, and timelines.
Add chapters to videos by transcribing, analyzing, and generating structured markdown documents with YouTube chapter markers. Optionally generate highlight videos.
AI4PKM helper for orchestrator configuration, worker management, and CLI workflow integration.
Convert Deckset-format markdown slides with speaker notes to presentation video with TTS narration. Use when user requests to create video from slides, generate presentation video, or convert slides to MP4 format.
| name | video-full-process |
| description | Unified workflow combining video-clean and video-add-chapters with transcript reuse and chapter remapping |
| allowed-tools | ["Read","Write","Bash","Glob"] |
| license | MIT |
Combines video-clean and video-add-chapters into a single workflow that:
# Full processing with default settings
python process_video.py "video.mp4" --language ko
# Preview mode (see what will be changed)
python process_video.py "video.mp4" --preview
# Skip chapter embedding (only clean + generate docs)
python process_video.py "video.mp4" --no-embed-chapters
Input: raw_video.mp4
│
▼
┌─────────────────────────────────────┐
│ 1. Transcribe (once) │
│ - Uses chunked transcription │
│ - Handles long videos (15min+) │
│ - Output: transcript.json │
└─────────────────────────────────────┘
│
┌─────┴─────┐
▼ ▼
┌─────────┐ ┌─────────────┐
│ 2a. │ │ 2b. Detect │ ← Parallel
│ Clean │ │ Chapters │
└─────────┘ └─────────────┘
│ │
▼ ▼
┌─────────┐ ┌─────────────┐
│ cleaned │ │ chapters. │
│ .mp4 │ │ json │
│ + pauses│ │ │
│ .json │ │ │
└─────────┘ └─────────────┘
│
▼
┌─────────────────────────────────────┐
│ 3. Remap chapters to cleaned video │
│ - Calculate removed time │
│ - Adjust chapter timestamps │
└─────────────────────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ 4. Embed chapters + Generate docs │
│ - ffmpeg metadata embed │
│ - YouTube chapters txt │
│ - Chapter markdown files │
└─────────────────────────────────────┘
│
▼
Output: cleaned-chapters.mp4 + docs
| File | Description |
|---|---|
{video} - cleaned.mp4 | Video with pauses/fillers removed |
{video} - cleaned-chapters.mp4 | Cleaned video with embedded chapters |
{video} - pauses.json | Removed pause data (for remapping) |
{video} - chapters.json | Detected chapter boundaries |
{video} - chapters_remapped.json | Chapters adjusted for cleaned video |
{video} - youtube_chapters.txt | Copy-paste for YouTube description |
Chapter NN - Title.md | Per-chapter documentation |
pip install openai
OPENAI_API_KEY - Required for Whisper APIpython process_video.py "presentation.mp4" --language ko
This runs:
# Skip transcription (reuse existing)
python process_video.py "video.mp4" --skip-transcribe
# Skip cleaning (only add chapters)
python process_video.py "video.mp4" --skip-clean
# Skip chapter embedding (only clean + generate docs)
python process_video.py "video.mp4" --no-embed-chapters
# Adjust pause threshold
python process_video.py "video.mp4" --pause-threshold 0.8
# Custom output directory
python process_video.py "video.mp4" --output-dir "./output"
When pauses are removed, chapter timestamps must be adjusted:
Original Video:
|--Ch1--|--pause--|--Ch2--|--pause--|--Ch3--|
0 30 40 60 70 90
Cleaned Video (pauses removed):
|--Ch1--|--Ch2--|--Ch3--|
0 30 50 70
Remapping:
- Ch1: 0 → 0 (no change)
- Ch2: 40 → 30 (10s pause removed before)
- Ch3: 70 → 50 (20s total pauses removed before)
The remap_chapters.py script handles this automatically using the pause data from video cleaning.
This skill orchestrates two existing skills:
transcribe_video.py, edit_video_remove_pauses.py--output-pauses flagtranscribe_video.py, suggest_chapters.py, generate_docs.py--skip-if-exists flagRun transcription first or use --force-transcribe:
python process_video.py "video.mp4" --force-transcribe
Regenerate remapped chapters:
python remap_chapters.py "video - chapters.json" --pauses "video - pauses.json"
Check that chapters.json format is correct:
{
"chapters": [
{"start": 0, "title": "Intro", "description": "..."},
{"start": 120, "title": "Main", "description": "..."}
]
}
video-full-process/
├── SKILL.md # This documentation
├── process_video.py # Main orchestrator script
└── remap_chapters.py # Chapter timestamp remapping