Mainly interested in Gamedev.
Looking for work.
Caution: May reblog sfw from 18+ accounts.
Breatharians DNI.


tdstr
@tdstr

for context: mvsep is a stem-separation AI website. you upload a song and it generates acapella/instrumental/etc stems. very useful for mashups/plunderphonics


went onto mvsep and saw 100 things in the queue at 5am, so i said fuck it i'll figure out UVR

good news despite being on github its actually very easy to set up, literally just an exe download. and i already used mdx-b which is just on there so not even a change

i dont know how it is for lower-end graphics cards but it only takes like 40 seconds for a 5 minute song on an rtx 3060 Ti, way faster than mvsep lol

all the youtube tutorials for it suck, just:

  1. go here https://github.com/Anjok07/ultimatevocalremovergui
  2. scroll down to "main download link"
  3. install and launch
  4. set "choose process method" to "MDX-Net"
  5. set "choose mdx-net model" to the one option there

and there you go, same settings as MDX-B

maybe its still a little taboo to mention mvsep publicly because its usability hinges on no one knowing it exists, but i think that ship's sailed unfortunately, it's been like a month of 60 things in queue minimum around the clock. just use UVR

originally posted on february 11, 2023


You must log in to comment.

in reply to @tdstr's post:

i used uvr fairly recently and i have a lower end graphics card and it took several minutes for me iirc but nothing too bad, and the results were better than what other stem separation things have given me anyway