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:
- Send a
hermes/asr/startListening
message with thestopOnSilence
property set totrue
- Send one or more
hermes/audioServer/<siteId>/audioFrame
messages with your voice command WAV audio - 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.