Tutorials

[Pants]

This application is quite impressive! Playing around with it has helped keep me sane during the pandemic, at least so far.

I do have a question/comment: I've noticed that it doesn't seem to be able to remove lower range male vocals all that well. Any idea if this is due to the training or if there's a setting I can change to adjust the low end floor for vocal removing? If it's a training issue, is it possible to request "low male vocals" as one of the future models? I'd attempt to train it myself, but my computer isn't quite powerful enough.

If not, no worries, I really appreciate what you've done with it so far and I'm looking forward to further updates.

[misterbeatzit]

So how exactly do I run the vocalremover.py file for the GUI? I tried "python vocalremover.py" but it gave an error.

[ChrisCall]

I do have a question/comment: I've noticed that it doesn't seem to be able to remove lower range male vocals all that well.

I second this. If it can be solved it would be massive. Lots of bassy voices that currently don't convert well enough, or parts of songs. Like Opeth isn't there yet, or The Tea Party, Cannibal Corpse forget about it lol.

On a more kind of mixed side you've got Pearl Jam. Lots of songs convert fantastic, but he tends to have this real low bass tone in some songs that never gets picked up.

[NewAgeRipper]

KaraokeDuov2 is this what you were talking about?

Yes sir. Awesome stuff.

[NewAgeRipper]

The Vocal Remover has been updated to v3.0.0: https://github.com/tsurumeso/vocal-remover/releases

Updated now to v3.0.1

[Anjok]

***UPDATE***

Hey guys! Sorry I've been MIA for awhile. There's been a lot going on and finding time to jump back into this has been difficult. I've successfully been able to get the GUI to work exactly as intended with v3.0. However, training for v3.0 has been another story.. I'm actually in the process of carefully going through my datasets to make sure they contain only the most effective pairs. The cleaner the dataset, the cleaner the vocal removal will be. My goal is to make it so the next model doesn't strip ANY instrumentation whatsoever.

An application update is on the horizon!

[Danizm]

hey mate will that mean that the acapellas will be cleaner to?

[Anjok]

hey mate will that mean that the acapellas will be cleaner to?

If you mean cleaner in that there's less instrumental bleed through, yes much cleaner. However, if you mean clean as in having the same overall sound quality as the generated instrumental, then no. Not yet. The issue isn't the AI either. It's because I don't have enough official acapellas to use in a dataset.

The way to train this AI to extract vocals with retained quality would be to add the official acapellas to the "instruments" folder and the mixes to the "mixture" folder. I actually coded a version of the GUI as "Ultimate Vocal Extractor". I just need a bigger dataset to train with, so it's still a work in progress.

[NewAgeRipper]

If you mean cleaner in that there's less instrumental bleed through, yes much cleaner. However, if you mean clean as in having the same overall sound quality as the generated instrumental, then no. Not yet. The issue isn't the AI either. It's because I don't have enough official acapellas to use in a dataset.

The way to train this AI to extract vocals with retained quality would be to add the official acapellas to the "instruments" folder and the mixes to the "mixture" folder. I actually coded a version of the GUI as "Ultimate Vocal Extractor". I just need a bigger dataset to train with, so it's still a work in progress.

Awesome news. Can't wait.

[Danizm]

Get this man some studio acapellas!!! how many would you need?