KVR Audio

metamorphosis · Post by **metamorphosis** » Wed Apr 12, 2023 7:18 am

Dunno if this's been mentioned here, but:
https://github.com/Anjok07/ultimatevocalremovergui

The results are apparently pretty stunning. I've got to head into the studio to download the 1.6gb installer.

Aloysius · Post by **Aloysius** » Wed Apr 12, 2023 10:06 am

metamorphosis wrote: Wed Apr 12, 2023 7:18 am I've got to head into the studio to download the 1.6gb installer.

harvon · Post by **harvon** » Wed Apr 12, 2023 10:33 am

While there're online tools for vocal removal too, this is interesting nevertheless.
I may have to change the PC however, the C: drive is tight, IK amplitube don't work, NI guitar rig don't work...

jamcat · Post by **jamcat** » Wed Apr 12, 2023 10:35 am

450mb download that expands to 1gb installed for Apple Silicon.

UVR left the instrumental faintly in the background of a track I tested, while gaudiolab didn't. However, it sounds like UVR does a little better at minimising the frequency loss in the separated instrumental track. It also seems to leave a little bit less residual instrumental colouration in vocals transients than gaudiolab.

metamorphosis · Post by **metamorphosis** » Fri Apr 14, 2023 12:33 am

Aloysius wrote: Wed Apr 12, 2023 10:06 am
metamorphosis wrote: Wed Apr 12, 2023 7:18 am I've got to head into the studio to download the 1.6gb installer.

Yeah no.
www.soulstudios.co.nz

metamorphosis · Post by **metamorphosis** » Fri Apr 14, 2023 12:35 am

jamcat wrote: Wed Apr 12, 2023 10:35 am 450mb download that expands to 1gb installed for Apple Silicon.

UVR left the instrumental faintly in the background of a track I tested, while gaudiolab didn't. However, it sounds like UVR does a little better at minimising the frequency loss in the separated instrumental track. It also seems to leave a little bit less residual instrumental colouration in vocals transients than gaudiolab.

Nice to know.
There's also spleeter for more generic instrument separation - which is used internally by some of izotope's stuff, if I understand correctly..

quolck · Post by **quolck** » Wed Jun 21, 2023 8:18 am

It looks like UVR uses the Demux model; spleeter is Deezer's inhouse model. I've liked spleeter for a while because it's open source, it's got the max/msp plugin for Ableton, and if you can get it running on a nice enough machine you can produce results quickly.

My problem is that I notice a lot of thinness and artifacts in my spleeter results. Something I thought of was playing around with comping+layering outputs from multiple runs on a song. Has anyone done this already or otherwise found ways to generate less lo-res sounding vocals from machine learning separation techniques?

jaydog4 · Post by **jaydog4** » Sun Jun 25, 2023 5:22 pm

https://hitnmix.com/

Now this Vocal stem separation tool I've found to be really cool

AI-based vocal removal tool - Ultimate Vocal Remover