I’m interested in being able to ask Rhasspy to take general dictation (e.g. by starting a dictation session), without losing the advantage of the fast STT system for recognizing the limited set of intents the rest of the time.
I see there’s a
speech_to_text.kaldi.open_transcription flag in the profile that would allow me to swap out the models when I start up Rhasspy, but I don’t see a way to do this per-message basis using the endpoints specified in
One way to do this might be to spin up a second instance of Rhasspy with only the Kaldi module running, but that seems silly. Is there an easier way to do this that I’m missing?