Validate and improve

[thewh1teagle]

I would like to use espeak-ng phenomenizer engine to extract phonemes in Hebrew, for using with Vits text to speech model.

Before I'm going to train the model I need to make sure that espeak extract it correctly since hevrew is complex language.

How can I check it?
And if it's wrong in some cases, how should I improve espeak-ng?

Thanks

[thewh1teagle]

Answer if someone needs that information in future:
Compile espeak and execute it with the input text and the right language flag, listen to the pronunciation and ensure it sounds well.
Edit the id_rules files where id is your language. It's simple English letters that is internal IPA representation of espeak, you should learn it fast.