search

metavoiceio / metavoice-src

Foundational model for human-like, expressive TTS
https://themetavoice.xyz/
Apache License 2.0
3.49k stars 615 forks source link

Phoneme-level pronunciation control #94

Open danablend opened 3 months ago

danablend commented 3 months ago

Hey! I understand that the text tokens are currently encoded on the character-level and the model is trained with these tokens.

What would be the process for getting phoneme level control over the output audio to correct pronunciations for exotic words or different accents during runtime? One could maybe fine tune models for this, but getting the phoneme level control on the input side would be great.

This would be an amazing add. Would be happy to contribute.

vatsalaggarwal commented 3 months ago

Where have you had these kinds of issues? Are you able to share examples?

StephennFernandes commented 2 months ago

@danablend

I get what you are trying to express. you should try dict TTS

It used a dictionary of pronounciation of exotic words with could of sentence as context. It's a context aware exotic word pronounciation model.