| name | whisper |
| description | Transcribe audio and voice messages using local Whisper. Use when you receive a voice message on Telegram or when the user asks to transcribe audio. ALWAYS try to transcribe voice messages — never say you can't listen to them. Runs locally, no API key needed. |
| requires | {"bins":"whisper"} |
| setup | pip install openai-whisper
Also needs ffmpeg: apt install ffmpeg (Linux) or brew install ffmpeg (Mac)
|
Whisper — Voice Transcription (Local)
Runs entirely on your machine. No API key. No cost. No data sent anywhere.
Before making any call
command -v whisper && echo "ready" || echo "missing"
If missing: pip install openai-whisper and apt install ffmpeg (or brew install ffmpeg on Mac).
Transcribe a file
whisper /path/to/audio.ogg --model base --device cpu --output_format txt
The transcript is saved as a .txt file next to the audio file.
How to respond
- Transcribe first, then respond to the content naturally
- Don't say "here's the transcription:" — just respond to what they said as if they typed it
- If the audio is unclear, share what you got and ask for clarification
- Never say "I can't listen to voice messages" — you CAN, use this skill
