Sending text data to TorToiSe TTS

[vandorb12]

I've been experimenting with AI voice cloning and voice generation for a little bit, and I would like to see how feasible it would be to get EDDI to hook into TorToiSe TTS https://github.com/neonbjb/tortoise-tts

I see that EDDI already has a "Write text copies of speech" function, and TorToiSe does have the ability to read from text files and also hook in with an API.

Currently at my skill level, I cannot get TorToiSe to actively monitor the *.out file that EDDI generates, thus making real(ish) time AI voice synthesis a little difficult to do. I'm sure with a little python scripting asking the computer to poll the *.out file for any changes and pushing it through TorToiSe would be a sufficient hack.

Just generally curious if anybody has developed any working solutions and is willing to show of examples.

[Tkael]

Not that I'm aware of.
I've tested a bit with https://replicate.com/afiaka87/tortoise-tts and while the end result is impressive the speech generation is a very long process and much too slow for an application like EDDI. Have you found some settings that are more optimal for real-time speech generation? Or it just a slow website?