Command Listener

As of Rhasspy 2.5, the speech to text system is responsible for detecting the boundaries of a voice command (the default systems use the rhasspy-silence library). Previously, this was done by a seperate "command listener" system. This page and its sections are here to avoid broken links and provide an explanation.

MQTT/Hermes

Rhasspy listens for messages according to the Hermes protocol to decide when to record voice commands. See the speech to text page for more details.

WebRTCVAD

The rhasspy-silence library used by Rhasspy's pocketsphinx and kaldi use webrtcvad to detect speech and silence.

OneShot

Deprecated as of Rhasspy 2.5

This system previously listened for a single WAV audio chunk and processed it as a complete voice command. You can acheive the same thing now with the following steps:

  1. Send a hermes/asr/startListening message with the stopOnSilence property set to true
  2. Send one or more hermes/audioServer/<siteId>/audioFrame messages with your voice command WAV audio
  3. Send a hermes/asr/stopListening message

With stopOnSilence set in startListening, the configured speech to text system should not attempt a transcription until the stopListening message is received.

Command

Deprecated as of Rhasspy 2.5

This system previously allowed for an external program to determine voice command boundaries.

Dummy

Deprecated as of Rhasspy 2.5

This system previously disabled voice command recording.