Closed mdarshad928 closed 1 year ago
@mdarshad928 Thanks for the contribution. In the meanwhile since I added Coqui support, Coqui introduced an improved Python API which can be seen here: https://tts.readthedocs.io/en/latest/#python-api
That means we can get rid of all Coqui-related custom logic and use that API instead. Getting word boundaries might still be an issue though.
As I am too busy to tend to the library these days, would you be interested in helping out with this change?
@mdarshad928 Thanks for the contribution. In the meanwhile since I added Coqui support, Coqui introduced an improved Python API which can be seen here: https://tts.readthedocs.io/en/latest/#python-api
That means we can get rid of all Coqui-related custom logic and use that API instead. Getting word boundaries might still be an issue though.
As I am too busy to tend to the library these days, would you be interested in helping out with this change?
Yeah, Surely. Let me go through the docs and I would certainly do the changes if I can.
Btw thanks for approving the pull request.
@mdarshad928 I've checked and Coqui indeed does not have word boundary support. I've created an issue for it here: https://github.com/coqui-ai/TTS/issues/2593
I think it's better if we deprecate the old logic before this feature is implemented on Coqui's side, since we already have the fallback word boundary detection through Whisper.
So I'd say go ahead. Here are some things that I had in mind to do:
pip install manim-voiceover[coqui]
. You might need to do some trial and error with poetry.Feel free to create a PR and copy over this information there. Lmk if you have any questions.
Hello.
I have added a few lines of documentation for Coqui services.