UVR returns mono output

[rkfg]

It seems that the underlying library produces good stereo output after music/vocal separation but after that these results are merged into mono with ffmpeg in your code, there's hardcoded number of channels as ac=1 in lib/audio.py => load_audio. Is it possible to return the audio as is?

[SayanoAI]

Most of the UVR models actually use mono audio as input so it would just add more processing time if you were to chain them. I guess I could parameterize the channel count...

[rkfg]

That would be nice! I tried roformer and it returns stereo.