If you're looking for other languages, Piper has been around in this scene for much longer and they have open-source training code and a lot of models (they're ~60MB instead of 25MB but whatever...) https://huggingface.co/rhasspy/piper-voices/tree/main

Actually I found it irritating that the readme does not mention the language at all. I think it is not good practice to deduce it from the language of the readme itself. I would not like to have German language tts models with only a German readme...

I tried on some Japanese for the kicks of it, it reads... "Chinese letter chinese letter japanese letter chinese letter..." :D

But yeah, if it's like any of the others we'll likely see a different "model" per language down the line based on the same techniques

TTS is generally not multilingual. One might think a well-annotated phonetic descriptions of voices would suffice, but that's not quite how languages work nor how TTS work.

(but somehow LLMs handle multilingual input perfectly fine! that's a bit strange, if you think about that)

Yes. The FAQ says that multilingual capabilities are in the works.