transcribe

name	transcribe
description	Transcribe a YouTube video or local video/audio file using OpenAI Whisper. Use when the user asks to transcribe a video, get a transcript, convert a video to text, provides a YouTube URL, or has a local MP4/audio file to transcribe. Triggers on: transcribe this video, get transcript, transcribe youtube, what does this video say, convert video to text, transcribe this file.

Transcribe the video or audio at $ARGUMENTS.

Instructions

1. Run the transcription script:

   python3 ~/.claude/skills/transcribe/transcribe_video.py "$ARGUMENTS"
   

2. The script will:
   • For YouTube URLs: Download the audio using yt-dlp, then transcribe
   • For local files: Transcribe directly (supports mp4, mp3, wav, m4a, etc.)
   • Save a timestamped transcript to ~/content/transcripts/transcript_<id>.txt
3. After the script finishes, read the saved transcript file from ~/content/transcripts/ and present the full contents to the user.
4. If the script fails, check:
   • Is OPENAI_API_KEY set in the environment?
   • Is yt-dlp installed? (only needed for YouTube URLs)
   • For local files: Does the file exist? Is it under 25MB?
   • For YouTube: Is the URL valid and publicly accessible?