Separate multiple speaker's voice from a single audio file

[tired oasis]

Is there an automatic way to filter and separate multiple voices talking in a podcast

[eternal sigil]

demucs would be your best bet with diarization

[livid panther]

any progress @tired oasis ?

[lyric mirage]

How do I run it fron the cloud?

[fickle isle]

Ayo? @lyric mirage level 1 !!!

[eternal sigil]

you using runpod, vast or what?

[lyric mirage]

Found a way to do it through MVSEP

[fickle isle]

Ayo? @lyric mirage level 2 !!!

[tired oasis]

So my dataset quality was really low
So didn't get good results in diarization

[kind ravine]

wdym low quality? like parts where two/more people speaking together/overlapping? then cut them

[tired oasis]

background noise was too much
cleaning noise with ai messes up the voice
idk if it matters, the audio was not in english too

[kind ravine]

UVR denoise, or better izotope RX spectral denoise
(never touched adobe podcast enhance that is said to cause some artifacts iirc)

[tired oasis]

tried adobe's (sounds like noises from horror movies lol) and UVR denoise not that good
gonna try izotope RX spectral denoise

[fickle isle]

Ayo? @tired oasis level 1 !!!