YouTube Transcript

Fetches a YouTube video's title and full transcript by pulling captions via yt-dlp. Prefers manual English subtitles, falls back to auto-generated English.

Requirements

• yt-dlp on PATH (brew install yt-dlp)
• Python 3

Usage

python3 ~/.pi/agent/skills/youtube-transcript/fetch_transcript.py "<youtube_url>"

Output

Prints a JSON object to stdout:

{
  "title": "Video title",
  "transcript": "full transcript text as a single string"
}

Progress/info logs go to stderr. On failure (no English captions, network error, bad URL), the script exits non-zero with a message on stderr.

Notes

• Only English captions are attempted (en, en-US, en-GB, then any en*). Manual captions are preferred over auto-generated.
• Transcript is plain text with timing/formatting stripped — not timestamped.
• For non-English videos or videos with captions disabled, the script will fail; consider video_extract with a Gemini prompt as a fallback.