Oct 13, 2026, 8:19 PM@tink @jakobrosin I should give the voice stuff a try as well since my old Google speaker is getting more bugs every day and I already have a Home Assistant setup. What do you use for voice I/O, TTS and speech-to-text?
- Oct 13, 2026, 8:19 PM
Replies
- Jakob Rosinjakobrosin@universeodon.comOct 13, 2026, 8:37 PM
@bram @tink I’ve used the home assistant voice as a satellite microphone / speaker box, with a local whisper STT and Piper TTS running on, surprisingly an intel mac book pro. In my case, the LLM processing it all was in the cloud, but with more decent hardware you could make it fully local pretty easily. WIth llm being in the cloud but rest local, I could get a plug toggled within 4 ish seconds after finishing the voice command.
- Oct 13, 2026, 9:04 PM
@jakobrosin @tink Hmm, I'm running it on my home server with a modern AMD CPU and lots of cycles and RAM to spare, so local should be fine. However, not sure if that ovice hardware is still available. Would be nice to repurpose the perfectly fine hardware in Google/Echo/whatever smart speakers, but hey, that's how it goes
Léonie WatsontinkOct 14, 2026, 9:24 AM@jakobrosin @bram Do you know if it's possible to change the speech rate universally using Piper? My experiments so far have used the Google TTS that was preinstalled with the HA OS and there doesn't seem to be a way to make it speak faster - and it's making my teeth itch already because it's so slow.
- Jakob Rosinjakobrosin@universeodon.comOct 14, 2026, 3:15 PM
Oct 14, 2026, 3:18 PM@jakobrosin @tink @bram I use my own eleven labs voice, because I am a narcicist.