Skip to content

Voice Capture

Local STT only. No meetings, no cloud Whisper, no always-on mic.

Procedure

  1. Take a path to an audio file the user recorded (their voice).
  2. Transcribe:
make voice AUDIO=/path/to/clip.wav
  1. Run pkm-capture on the transcript with capture_source: voice. Classify first.
  2. Do not send audio or transcript through OmniRoute if classification is private_sensitive.

Anti-patterns

  • Don't record coworkers or meetings.
  • Don't call a cloud STT API.