What UVR model should I use to separate main and backing vocals?

[jaunty oriole]

I'm making an RVC cover of a song that consists of the main vocals and and a backing choir, nothing else. I want to cover only the main vocals, and leave the choir as is. Is there a UVR model that can do this?

[green plinth]

Use vocal remover HQ 3 or 5

[jaunty oriole]

those are meant for vocals and instrumentals

I've technically got only vocals

[green plinth]

yeah uvr-mdx-net hq 3, or number 5 they actually isolate the main vocals, sometimes backing vocals don't work because the ai can not fully reconigze the vocals with most songs i use it could but it's not perfect

maybe use x-minus vocal remover

[rocky hatch]

https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c/edit?tab=t.0#heading=h.vg1wnx1dc4g0

[polar inlet]

^search melroformer karaoke in the above docs

[steady grove]

Look for commercial audio plugins costs alot but can get it free... better than that tbh

[polar inlet]

afaik only dango.ai has great proprietary separation models, most other ones are just spleeter/demucs based garbage

[winter ore]

Dont use x minus wth roscall 😭

[polar inlet]

there's BVE v2 exclusively in x-minus

[winter ore]

Isnt that its only claim to fame and isnt it only very slightly better than normal bve

[polar inlet]

more options is better if there are some harmonies & adlibs hard to separate

[winter ore]

Isnt that its only claim to fame and isnt it only very slightly better than normal bve

More options are nice yea

I often run like 4 different karaoke models and patch them all together lmao

BVE is solid asf in terms of actually targeting second voices. Muddiness is worse than more modern stuff tho