🎬 Video-to-Notes Skill

Transform any YouTube/Bilibili video into structured AI-powered notes — just by chatting with Claude.

中文文档

---

✨ What This Does

Drop a video URL in Claude Code, and get back:

• Structured markdown notes with key points, chapters, and code examples
• Smart formatting for technical content (commands, code blocks, examples)
• Privacy-first processing - audio never leaves your machine
• Ultra-low cost - ~$0.02 per video (only AI summary costs money)

Perfect for: Tutorial videos, tech talks, lectures, documentation videos

---

🚀 Quick Setup

Step 1: Install Dependencies

System Requirements:

• FFmpeg - for audio extraction
• yt-dlp - pip install yt-dlp

Python Dependencies:

cd scripts
pip install -r requirements.txt

Step 2: Get Your API Key

1. Visit OpenRouter.ai
2. Sign up and create an API key
3. Add $5-10 credits (enough for 200+ videos)

Step 3: Configure

cp .env.example .env
# Edit .env and add: OPENROUTER_API_KEY=sk-or-your-key-here

That's it! The skill is ready to use.

---

💬 How to Use

In Claude Code (Recommended)

Just talk naturally to Claude:

English:

Convert this video to notes: https://www.youtube.com/watch?v=a9eR1xsfvHg

Summarize this tutorial in English: https://www.youtube.com/watch?v=...

Please save the notes to my Desktop/notes folder

中文:

请将这个视频转换为笔记：https://www.bilibili.com/video/BV1xx411c7XZ

帮我总结这个教程，用中文

把笔记保存到我的文档文件夹

Claude will:

1. ✅ Check your environment automatically
2. ✅ Download and process the video
3. ✅ Generate beautiful markdown notes
4. ✅ Save to file if you ask

---

📊 What You Get

See real example: examples/github-spec-kit-notes.md

Every note includes:

• Clear title based on video content
• Video metadata (source, duration, author, timestamp)
• Core takeaways summary (3-7 bullet points)
• Structured sections with hierarchical headings
• Technical terms formatted as inline code
• Code examples with syntax highlighting
• Smart examples when concepts are complex

---

🎯 Capabilities

Feature	Details
Platforms	YouTube, Bilibili, Xiaohongshu
Languages	Auto-detect, or specify: Chinese, English, Japanese, etc.
Video Length	Up to 2 hours (configurable)
Speed	~2-3 minutes for an 8-minute video
Cost	~$0.02-0.05 per video (AI summary only)
Privacy	Audio processed locally with Whisper

AI Models Available:

• google/gemini-2.5-flash (default - fast & cheap)
• anthropic/claude-3.5-sonnet (best quality)
• openai/gpt-4-turbo (balanced)

---

⚙️ Advanced Configuration

Environment Variables (Optional)

Edit .env to customize:

# Required
OPENROUTER_API_KEY=sk-or-your-key-here

# Optional - customize these if needed
AI_MODEL=google/gemini-2.5-flash    # Which AI to use
WHISPER_MODEL=base                   # Whisper model: tiny/base/small/medium/large
DEFAULT_LANGUAGE=zh                  # Default language: zh/en/auto
OUTPUT_DIRECTORY=.                   # Where to save notes

Whisper Models

Model	Speed	Accuracy	RAM	Best For
tiny	⚡⚡⚡	⭐⭐	1GB	Quick drafts
base	⚡⚡	⭐⭐⭐	1GB	Default - balanced
small	⚡	⭐⭐⭐⭐	2GB	Better accuracy
medium	🐌	⭐⭐⭐⭐⭐	5GB	High quality

---

🛠️ For Developers

Architecture

Video URL → yt-dlp → FFmpeg → Whisper (local) → OpenRouter API → Markdown

Command Line Usage

If you prefer running scripts directly:

python scripts/process_video.py \
  --url "https://www.youtube.com/watch?v=..." \
  --language zh \
  --save-to-file \
  --output-path "./notes"

---

📝 License

MIT License - See LICENSE

---

Built with yt-dlp · FFmpeg · OpenAI Whisper · OpenRouter

⭐ Star this repo if it helps you!