Nice. I hope they’re also going to release the training process and include the community. And my first language is missing. But a TTS model at that size would be awesome, especially if it sounds as good.

Ultimately I’d like to see something like this being adopted and tied into the desktop and maybe my phone. Because honestly, neither Linux nor the open source part of Android has a good and modern solution for TTS that is easy to use and hooks into other software.

It is even better: they are showing that even training data does not need to be huge.