I’ve just set up my new, first Voice Assistant PE device. All seems to work okay, except that most voice responses are broken up, as in:
“Sorry, I”…“couldn’t under”…“stand that.” or “Turned on the s”…“switch”.
The responses seem parsed in segments of three or four syllables, with each pause between segments at least a half second in duration.
It takes about five seconds for the voice responses to be delivered. A request to turn on a light took effect in about two seconds, but the response acknowledging it lagged by another two or three. And some voice responses are truncated, such as “Sorry, I”…“couldn’t” - and nothing more.
I’ve been using an RPi 4B-based Wyoming Voice satellite for the past year, which does not suffer any of this. For both devices, both Piper and Fast Whisper run locally.
OpenAI is a fee-based cloud thing, correct? If so, it’s not for me.
I’m glad you got your PE to work in that way, but I suspect the stuttering issue can’t be endemic to everyone employing Assist. Hopefully there’s a fix for PE w/ Assist users experiencing this issue.
Yes, I have a few credits on OpenAI, so yes it is pay-to use and it is not my wish solution as well. I was testing it in fall back mode, so, in theory as per the 2024.12 announcement, I expected to send commands to the voice assistant and the basic commants were handled locally by assist, using OpenAI only when Assist cannot respond. It ended up with Assist never working to me with that stuttering issue you mentioned.