Kokoro is an open-weight TTS model (82M parameters). Its lightweight and delivers comparable quality to larger models. With Apache-licensed weights, it can be deployed anywhere! I think it would be great if someone could make a plug in... It is beyond me - but would be great to have something other than the Piper read aloud.
Is this a good idea? Possible?
https://github.com/hexgrad/kokoro?tab=readme-ov-file
https://github.com/Lyrcaxis/KokoroSharp
