Local tts/stt integration

, ,

Hello @all,

I am looking for an integration for local speech-to-text/text-to-speech processing in my home network, but separate from HA. For now, I’ve tried the MS Azure integrations for both and they work great - but I’d like to explore doing it locally. I’ve created a local LocalAI VM - they’re offering some tts/stt modules as well, but I could not find any integration to send the data to my local VM(s). Maybe it can also be configured manually, but as I am just beginning using HA, I am missing the spot to begin here. In this case I also would have to configure the JSON requests and responses, I guess.

Any ideas? Thanks in advance.

Have you seen this?

It used to be called faster whisperer, an implemention of the open -ai whisperer stuff outside of HA and can use a GPU.

You can use wyoming faster-whisper and piper on any machine - doesn’t have to be a HA Add-on.

I personally use wyoming-faster-whisper and kokoro-fastapi for STT and TTS respectively. These run on a separate machine.

I haven’t been able to figure out how to connect a container running kokoro-fastapi for TTS with home assistant. How did you do that? Any hints or docs you can share?

After posting this I came up with a better way to search and figure this out. For anyone who is also searching i was able to use a HACs custom component and set the openai endpoint to my Kokoro-fastapi service and it worked great.
I am still interested if others have found a different solution, but this worked well for me as of March 2025.

1 Like