Skip to content

Voice Memos

Clawbolt supports voice messages in Telegram. Send a voice note and it gets transcribed and processed as if you typed the message.

  1. Record a voice note: Use Telegram’s built-in voice message feature
  2. Transcription: The audio is processed by faster-whisper, a fast and accurate speech-to-text engine
  3. Processing: The transcribed text is fed into the agent loop just like a typed message
  4. Response: Clawbolt responds to the content of your voice note

Voice memos are great for:

  • On the job site: When your hands are dirty or you’re holding tools
  • Driving between jobs: Quickly capture thoughts or requests
  • Detailed descriptions: When typing would take too long
  • Estimate requests: Describe the job verbally and get a PDF estimate back

Voice transcription requires ffmpeg for audio processing. The Docker image includes ffmpeg by default. For local development, install it via your system package manager.