vocal removal tool to remove everything but the vocals?

[gavilatius]

I was thinking can you use the vocal removal to remove everything but the vocals? I want to keep the vocals but get rid of everything else. I'm asking this question because I'm doing a dubstep of Pinkie Pie's "no fear song" from My Little Pony: Friendship in Magical, and I don't want the original background music to offset the new tempo.

here's the original song BTW: http://www.youtube.com/watch?v=-tiynTPu ... detailpage

[DVDdoug]

First, "vocal removal" is generally not very satisfactory. It works by simply subtracting the left from right, and it removes everything in the "center". If the vocals are not equal (and in-phase) in both channels, the vocals won't be completely removed.

I want to keep the vocals but get rid of everything else.

This is a bit trickier, but I know of a couple of programs that attempt it:
DSP Centercut (plug-in for Winamp)
GoldWave has a Stereo Center effect.

I've done a little experimentation with these, and again the results are "interesting", but not generally satisfactory.

You can't un-fry an egg or unbake a cake, and you can't un'mix a song!

[kozikowski]

That's called "Vocal Isolation" for your Googling pleasure.

All these tools work on a Stereo show. If the original isn't in Stereo -- really good stereo, no MP3 downloads -- then you might have a small chance.

This is one of those tools whose description sounds really cool on paper, but rarely works well. Noise Reduction is the other one.

We get killed regularly by someone on YouTube successfully creating a karaoke track demonstration and sucking the vocal out of a song leaving just the music behind. They picked the one perfect song where that works very well with no artifacts, damage or leakage. Most songs don't work like that.

Koz

[gavilatius]

if thats the case is there a way to increase the vocals while decreasing the background music?

[kozikowski]

No, because its the same problem. The software needs to "know" which sounds are the voice. Past the usual trick that voices are usually dead center in a stereo performance, there isn't a way. Large pipe organs have a stop called "Vox Humana." It makes the organ sound like a small human chorus. If you close your eyes, it's rough to tell which is which.

The software needs a "hook" or some way to recognize the voices. Most performances have no hook.

It's common to try to do this with equalizer tools -- separation based on pitch of the notes. The problem there is the overtones and richness of the notes tends to be common. The same thing that makes a voice rich and pleasant also makes a violin playing the same notes rich and pleasant.

Somebody on the fourth floor once said they had software that would do this and I got all excited about it. Turns out they could only do it one note at a time -- and even then it didn't work all that well.

Koz