I implemented a ready to use gradio app with more readable code and locally run models

[huynhnhathao]

For anyone who is interested, I have implemented a Gradio UI app of BARK that is ready to run on your local computer, plus most of the code in this codebase is reimplemented to be more readable
I also trained a HuBERT model to predict the semantic tokens from audio from a more than 4700 generated ~14s examples of audio-semantic dataset, the validation accuracy of my model is 83% on more than 15k tokens, one can get a feeling of cloning a voice but it is not perfect to impersonate anyone, plus you can do all that in the UI, touching no code

https://github.com/huynhnhathao/bark_text_to_audio