Short demo of custom voice agent | Home Assistant | Page 1

hexed pewter Dec 20, 2025, 3:22 PM

#

STT is faster whisper.
LLM is currently Grok-4-fast as it has so far be the best combo of need d intelligence to tool call correctly and speed.

spiral hawk Dec 20, 2025, 4:55 PM

#

would have posted that in #1284960456880029707 ..

wooden pagoda Dec 21, 2025, 11:49 PM

#

how are you running the voice model?

hexed pewter Dec 22, 2025, 5:04 PM

#

I'm going to delete this, I didn't realize voice-assistants was a sub-channel of "Support", which this is not.

@wooden pagoda I'm using a fork of chatterbox-tts-api that I hacked to add the newer Chatterbox Turbo model, and then using OpenAI TTS to point to that running on the laptop

#Short demo of custom voice agent