Data produced by the microphone module
The module’s own README describes an earlier design: a CommsEvent with Content: "active" /
"inactive" marking utterance start and end, and a Secure flag on every event reflecting
exclusive-mode. None of that matches the current source. What follows is verified against the
module’s actual code, not the README.
Entity
One entity per session, id derived from a hash of the capture device’s own WASAPI device ID string.
Utterance start and end
Marked with a TrackingEvent - a payload-free,
one-shot occurrence record, not a CommsEvent - of type VOICE_TRANSMISSION_STARTED or
VOICE_TRANSMISSION_ENDED. Neither carries a Secure flag; TrackingEvent has no such field.
An utterance starts the instant the peak level crosses SilenceThresholdDb, and ends only after
SilenceGapMs of continuous silence following it - so the end event is always published later
than the moment speech actually stopped, by design.
Audio content
Only published when EnableAudioInComms is on, and only if the utterance actually produced
audio bytes (which requires a capture stream to have been running - see
Configure). One
CommsEvent per source, comms_type: COMMS_TYPE_VOICE, published
when the utterance ends:
| Field | Value |
|---|---|
target | Name of the process holding the active audio session at the moment speech was detected, empty if none could be identified |
content | audio/pcm;src=mic;rate=RATE;bits=BITS;ch=CHANNELS;enc=base64,DATA (actual sample rate, bit depth and channel count in place of the capitalized names) for the microphone, or the same with src=desktop for the loopback stream in MicPlusDesktopSeparated mode |
secure | Not set (absent) on this event, regardless of exclusive mode |
With nothing to attach - EnableAudioInComms off, or no capture stream was running - no
CommsEvent is published for the utterance at all; only the start/end TrackingEvents are.
Transcript
Only published when EnableTranscription is on and the utterance’s microphone audio produced a
non-empty transcript. One CommsEvent, published alongside (not instead of) the audio content
event above if both are enabled:
| Field | Value |
|---|---|
comms_type | COMMS_TYPE_TTS - reusing the text-to-speech enum value for this speech-to-text transcript, not COMMS_TYPE_VOICE |
target | Same active-process name as the audio content event |
content | The transcribed text |
secure | The implementation sets this from exclusive-mode state. Exclusive mode disables microphone capture, so this release cannot produce a transcript event with secure: true. |
Transcription runs only against the microphone’s own audio, never the desktop loopback stream,
even in MicPlusDesktopSeparated mode.
What is never published
Peak/RMS level is not published in any form - it exists only inside this module’s own debug
view. There is no BiometricSensorComponent reading anywhere in this module’s output, regardless
of settings.