Feature/speech

[HarrisonFisher]

No description provided.

[rubens-soto]

Kokoro TTS is a good option to run locally on cpu as well:
https://github.com/hexgrad/kokoro

[HarrisonFisher]

@rubens-soto The audio samples sound great — really impressive for a model under 500 MB. Does it run fast on just a CPU? 😮

Since I was able to write a script and use edge-tts via v1/code/execute/python, I was planning on closing this PR. 🤷‍♂️

I’m definitely interested in building something around Kokoro (or similar lightweight models). I’m just not sure where @stephengpope stands on adding another model to this API. If he’s open to it, I’d be happy to explore this. If not, I’ll probably hold off for now, since I think some things in motion will make it easier to approach this down the line in a separate project. Appreciate you surfacing it!

[rubenssoto]

Yes, it does. I used to run locally in a ryzen 5300.