youtube-summarizer

YouTube Summary

Fetches a YouTube video transcript, generates a concise summary with key takeaways and tag suggestions, and saves a structured note to the Obsidian vault's REFERENCES/ folder with the raw transcript in TRANSCRIPTS/.

When to Use

Invoke /youtube-summarizer <url> when you want to:

• Capture a YouTube video as a reference note in your vault
• Get a concise summary and key takeaways without watching the full video
• Build a searchable library of video content

Prerequisites

Install required Python packages:

pip install youtube-transcript-api yt-dlp

Or use uv (no pre-install needed):

uv run --with youtube-transcript-api --with yt-dlp skills/youtube-summarizer/fetch_youtube.py ...

Workflow

1. Discover Vault Root

Read Obsidian configuration to find the active vault:

cat ~/Library/Application\ Support/obsidian/obsidian.json

Parse the JSON to find the vault with "open": true, or use the most recently opened vault. Extract vault_root from the path field.

2. Fetch Transcript and Metadata

Run the fetch script:

python3 skills/youtube-summarizer/fetch_youtube.py "<vault_root>" "<youtube_url>"

The script outputs JSON to stdout:

{
  "video_id": "dQw4w9WgXcQ",
  "title": "Video Title",
  "channel": "Channel Name",
  "duration": "3:32",
  "published": "2009-10-25",
  "note_path": "/vault/REFERENCES/2009/10-October/2009-10-25 - Video Title.md",
  "transcript_path": "/vault/TRANSCRIPTS/2009-10-25 - Video Title - transcript.md",
  "timestamped_transcript_path": "/vault/TRANSCRIPTS/2009-10-25 - Video Title - transcript-timestamped.md",
  "heatmap_path": "/vault/TRANSCRIPTS/2009-10-25 - Video Title - heatmap.md",
  "already_exists": false,
  "transcript_length": 4821
}

timestamped_transcript_path will be null if no transcript is available. heatmap_path will be null if the video has no YouTube heatmap data (requires sufficient view count).

3. Handle Already-Existing Notes