Questions about the features of the history prompt

[susenyang]

I want to use my audio as the prompt, but I need to extract the corresponding semantic features, coarse features and fine features. In order to align the pre-trained models, can you tell me which version of models(hubert_kmeans, Encodec) you are using to extract features?

[asr-pub]

Use this fork: https://github.com/serp-ai/bark-with-voice-clone/blob/main/clone_voice.ipynb

[gkucsko]

unfortunately we can't release that part to prevent unauthorized voice cloning. there are some forks with approximations like the one above (not maintained by suno) that you can have a look at. But probably the cloning won't sound particularly similar to the prompt in most cases.