Supporting other languages

[alloystorm]

Do I understand it correctly that since the phonetization part is espeak-ng, which supports a lot more languages. It is just about training additional models to allow Piper to support other languages?

For example adding support for Japanese and Korean.

Also I'm a bit unclear whether improvements in natural pauses, intonation, and speed variations can be done in the VITS models (by training new models with better training data) or are these areas largely limited by espeak-ng?