French model released

[RASPIAUDIO]

Checks

• [X] This template is only for question, not feature requests or bug reports.
• [X] I have thoroughly reviewed the project documentation and read the related paper(s).
• [X] I have searched for existing issues, including closed ones, no similar questions.
• [X] I confirm that I am using English to submit this report in order to facilitate communication.

Question details

Hi Please find an improvised tutorial on how to train a new language for F5-TTS simply using the gradio web interface

https://www.youtube.com/watch?v=UO4usaOojys

Find also the link for the french model I have trained on 80k samples, 100 epochs, on single speaker, the google drive link is in the video comment. As well as training material used in the video description. Please subscribe to my channel to support!

My feeling is that the more I train the less sensitive it gets to the reference sample that I want to clone and that the result come closer to the training sample, any idea how to overcome this?

[SWivid]

Hi @RASPIAUDIO , will need diverse training corpus. It's normal case if only single speaker to train.

[RASPIAUDIO]

thanks I will try to complete the training with many different speakers