AU
audio-transcription
Transcribe inbound audio clips to text using local Whisper (faster-whisper). Use when the user sends voice notes/audio files and asks what was said, asks for transcript/summary/action items, or asks to process audio privately on-host. Default model is small.en for better English accuracy on constrai
Install
mkdir -p .claude/skills/audio-transcription && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16329" && unzip -o skill.zip -d .claude/skills/audio-transcription && rm skill.zipInstalls to .claude/skills/audio-transcription
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Transcribe inbound audio clips to text using local Whisper (faster-whisper). Use when the user sends voice notes/audio files and asks what was said, asks for transcript/summary/action items, or asks to process audio privately on-host. Default model is small.en for better English accuracy on constrained hosts.310 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
About this skill
Audio Transcription (Local, small.en)
Use local transcription by default with small.en.
Default command
python3 skills/audio-transcription/scripts/transcribe_local.py \
--input "<audio-path>" \
--model small.en
Fast/lower-resource fallback
python3 skills/audio-transcription/scripts/transcribe_local.py \
--input "<audio-path>" \
--model base.en
What to return to user
- Transcript (verbatim best-effort)
- Short summary (2-4 bullets)
- Action items/questions extracted from transcript
- If confidence seems low, say so clearly
Notes
- First run downloads model weights (~500MB for
small.en). - Keep transcripts concise in chat; offer full dump on request.
- Treat audio content as sensitive by default; do not share externally.