RVC Dataset

#3
by ardha27 - opened

hello, can you help me? i'm trying to use RVC colab version and it need dataset like this "/content/drive/MyDrive/dataset/lulu20230327_32k.zip". what type of dataset should i provide inside the zip?

image.png
how many epoch should i train it?

I am also testing. I recommend 200-500 epochs.

can you explain to me what is Retrieval feature ratio ? what does it mean if its set to 0,6?

The Retrieval feature ratio is the ratio of the combination of input audio features and training set features, used to reduce tone leakage and improve the timbre similarity. 0.6 is the recommended value.

Can i infer rvc model using sovits GUI?

Of course not

oww okay, thanks for the information

ardha27 changed discussion status to closed

Sign up or log in to comment