From gobbler
Transcribes audio and video files to markdown using Whisper. Triggers on .mp3, .wav, .m4a, .mp4, .mov files, or requests to transcribe podcasts, recordings, voice memos, interviews, or spoken content.
How this skill is triggered — by the user, by Claude, or both
Slash command
/gobbler:gobbler-audioThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Transcribe audio and video files to markdown using Whisper.
Transcribe audio and video files to markdown using Whisper.
Requires: ffmpeg installed, Whisper model (auto-downloads on first use)
# Basic transcription (outputs to stdout)
gobbler audio /path/to/audio.mp3
# Save to file
gobbler audio /path/to/recording.m4a -o transcript.md
# Choose model size (tiny, base, small, medium, large)
gobbler audio /path/to/video.mp4 --model medium
# Specify language (auto-detect by default)
gobbler audio /path/to/audio.wav --language en
# Include timestamps in output
gobbler audio /path/to/audio.mp3 --timestamps
# Choose output format (markdown, json, table)
gobbler audio /path/to/audio.mp3 --format json
# Use a different transcription provider
gobbler audio /path/to/audio.mp3 --provider whisper-local
| Option | Short | Description | Default |
|---|---|---|---|
--output | -o | Output file path (stdout if not specified) | - |
--language | -l | Audio language (auto-detect if not specified) | auto |
--model | -m | Whisper model size | small |
--timestamps | - | Include timestamps in output | no-timestamps |
--format | -f | Output format (markdown/json/table) | markdown |
--provider | -p | Transcription provider | whisper-local |
--skip-if-exists | - | Skip if output file already exists (batch-friendly) | - |
| Model | Speed | Accuracy | Memory |
|---|---|---|---|
| tiny | Fastest | Lower | ~1GB |
| base | Fast | Moderate | ~1GB |
| small | Moderate | Good (default) | ~2GB |
| medium | Slower | Better | ~5GB |
| large | Slowest | Best | ~10GB |
When saving transcripts to a file, follow these steps:
gobbler config get output.default_directory
If a default directory is configured, use it with a descriptive filename:
gobbler audio /path/to/recording.m4a -o "<default_directory>/Recording Transcript.md"
If the config returns empty/null, save to the current directory or ask the user where to save:
gobbler audio /path/to/recording.m4a -o "Recording Transcript.md"
-l en (or your audio's language) for 7x faster transcription! Auto-detection adds significant overhead. Example: 14s → 2s for a short clip.small model offers best speed/accuracy tradeoffGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub enablement-engineering/gobbler --plugin gobbler