XTTS for Persian

[alikhabazian]

I’d like to train or fine-tune XTTS for Persian.
Should I just run the code in the ljspeech folder, or is there a better way to set it up?
Do you have any recommendations for preparing and training with a Persian dataset?

Also, I have a few more questions:
• When training a new language, do we train the whole XTTS model or just the GPT part?
• Since Persian and Arabic share some similarities, is there already an Arabic-trained model available, and can I reuse that?
• For adding a completely new language, do I need to train the dVAE again from scratch?

[eginhard]

You can check https://github.com/anhnh2002/XTTSv2-Finetuning-for-New-Languages