Add feature

[Suhailyu]

Both male and female vocals are included in the conversion process, and the effects apply to both types of sounds. Any improvements made depend on the selected model. For example, if a male model is chosen for conversion, it will only affect male vocals, leaving female vocals unchanged and unaffected by the conversion

[JackismyShephard]

@Suhailyu This might be a useful feature but it is not easily attainable. The way song cover generation currently works is that first the main vocals are extracted from the selected song and then those vocals are converted using RVC. For what you are suggesting you would need a stemming model that separates an audio track into three stems: male vocals, female vocals and instrumentals. I am not sure there are stemming models that can do that well. Also you would need to know a priori whether a chosen voice model is based on a male or female voice. I don't think that information is available at the moment for all RVC voice models.

[Suhailyu]

I’ve noticed an issue where there’s leakage in some songs, and the conversion process is also affecting the instruments.

[JackismyShephard]

Yeah that is also due to the separation models not being perfect. You can try my fork of this app where you have the possibility to experiment with some settings for the audio separation models. It might help.