how to train from scratch for other language?
I'm interested in training this for a different language. Could you guide me on how to get started, including instructions and data preparation?
Hey.
This network isn't exactly easy to train for people new to the field. you may find it easier to start with other models, particularly those that don't train end to end.
still, I don't have a specific answer to your question; because each language has its own unique challenges (such as proper Phonemization).
I trained the PLBERT and ASR models from scratch. Now, I'm considering how to integrate these models with either the original StyleTTS2 repository or your repository. I'm particularly interested in your repository because of the excellent improvements you've made.
That's great to hear!
it should be a simple drag and drop. I tried my best to keep Tsukasa as close to the original's code base as possible. so you should be mostly good by simply replacing your pre-trained models with the ones here or there.