#What’s the most up-to-date way to hook assist up to OpenAI and stream TTS?

1 messages · Page 1 of 1 (latest)

dreamy walrus
#

I finally decided to hop on the voice train and hook my house up to an OpenAI LLM. It doesn’t seem like there are too many guides on how to do this and some seem pretty dated.

My current bodice pipeline is running in a Proxmox VM running HassOS and looks like this Whisper (STT) running as an add-on -> OpenAI Conversation Integration -> Wyoming OpenAI (docker container on a separate system) (TTS).

It works pretty well, but the TTS responses from Wyoming OpenAI don’t seem to be streaming. Instead, the logs make it appear that Wyoming OpenAI is synthesizing the entire text in one go and isn’t breaking it into chunks. Is this something that is currently supported?

Any advice on how to get TTS streaming would be very helpful! Also, if there’s a better way to set this all up, I’m open to it!

#

What’s the most up-to-date way to hook assist up to OpenAI and stream TTS?

tacit mountain