transcribe

Video Transcriber

Transcribe any video URL into text using yt-dlp (download) + faster-whisper (transcription).

Install

If the user says /transcribe install or /transcribe setup, run this install wizard instead of the normal workflow. Walk through each step conversationally — check, report, fix, move on.

Step 1: Check Python 3

python3 --version

Need Python 3.8+. If missing or too old, guide the user:

• macOS: brew install python3
• Linux: sudo apt install python3 python3-pip
• Windows: Download from python.org

Step 2: Check pip packages

pip3 show faster-whisper 2>/dev/null && echo "OK: faster-whisper" || echo "MISSING: faster-whisper"
pip3 show yt-dlp 2>/dev/null && echo "OK: yt-dlp" || echo "MISSING: yt-dlp"

For any missing packages:

pip3 install faster-whisper yt-dlp

Note: faster-whisper will download a ~1.5GB Whisper model on first use. This is normal and only happens once.

Step 3: Check ffmpeg

which ffmpeg && ffmpeg -version | head -1 || echo "MISSING: ffmpeg"

If missing:

• macOS: brew install ffmpeg
• Linux: sudo apt install ffmpeg
• Windows: Download from ffmpeg.org and add to PATH

Step 4: Verify the script

python3 -c "import faster_whisper; import yt_dlp; print('All Python dependencies OK')"

Step 5: Report status

Print a summary like this:

Transcribe — Install Summary

  Python:         [OK/MISSING] (version)
  faster-whisper: [OK/MISSING]
  yt-dlp:         [OK/MISSING]
  ffmpeg:         [OK/MISSING]
  Script:         [OK/MISSING]
  Status:         [Ready to use / X items need attention]

Try it: /transcribe https://youtube.com/watch?v=example

If anything failed, summarize what still needs fixing before the skill will work.

---

Invocation

/transcribe <url> — the URL is passed as the skill argument.

If no URL is provided in the argument, ask the user for it.

Workflow

1. Ask the user: "Include timestamps?" (Yes / No)
2. Run the transcription script:

python3 <skill_path>/scripts/transcribe_url.py "<url>" [--timestamps]

Add --timestamps if the user chose yes.

1. Present the transcript output to the user — the script prints the transcript directly to stdout.

Dependency Errors

If the script exits with missing dependency errors, relay the install instructions it prints:

pip3 install yt-dlp faster-whisper
brew install ffmpeg  # macOS

Options

• --model <size> — Whisper model size (default: medium). Use small for faster but less accurate results, large-v3 for maximum accuracy.
• --timestamps — Prefix each segment with [M:SS] timestamp.