Microphone module

The microphone module detects voice activity on a Windows audio capture device using WASAPI’s peak meter, and marks when speech starts and stops. Recording the actual audio, and transcribing it with a locally-run Whisper speech-to-text model, are both separate, off-by-default settings - with nothing enabled beyond the defaults, this module never opens an audio capture stream at all, only the lightweight peak meter.

Privacy

Turning on EnableAudioInComms or EnableTranscription means this module records and stores the content of what was said during a session, not only that speech occurred. Treat that as you would any other audio recording of a session: get the necessary consent for your deployment before enabling either. See Configure and Data produced for exactly what is captured under each setting.

What you need

• A Windows audio capture device (any microphone Windows recognizes over WASAPI).
• EnableTranscription needs the bundled Whisper GGML model. It ships inside the module package; see Troubleshooting if it is reported missing.

Interface Control Document

See Interface control document for WASAPI formats, voice-activity and utterance lifecycle, frame schemas, transcription dependencies and privacy boundaries.

Pages

• Configure: capture mode, voice-activity thresholds, and the audio/transcription settings.
• Data produced: what is actually published, corrected against a README that describes an earlier design.
• Troubleshooting: symptoms, causes and fixes.

HEAT Capture team. Last updated 2 October 2026.