how to remove music and keep vocals?

[Parbruek]

To do the mathematical opposite of what you do when you remove vocals, you would need to have a program which retained the lowest bit above the center line (or highest below it). But Left and Right waves are not universally on the same side of the center line, so you would have to keep the line of "Mid" on the same side, whenever the wave on the other side of the center line became shorter than the wave which was being followed.
Therefore from such a program you would receive only a distance from the center line as dependable, but the side on which the wave was would just have to alternate. So what you would get might sound like Mid extracted from stereo, but problems would arise if you had to mix it back together with, say the Left-Right or Right-Left combination.

But there must be other problems though, otherwise someone would have written the program, right?

[kozikowski]

Oh they have...

--Voice Trap
http://www.cloneensemble.com/vt_main.htm
--Extra Boy
http://www.kvraudio.com/get/1651.html
--Kn0ck0ut
http://www.freewebs.com/st3pan0va/


It's called Vocal Isolation and there are a couple of money-based products that claim to be able to do that. One of the packages also claims that the singer doesn't have to be in the stereo middle which is a common problem with free programs.

Koz

[Thunderbolt1000T]

I have a similar problem, but I'm trying to isolate the center channel, and anything that may be on it. I've tried and tried copying and pasting and using the invert effect, and I'm still getting nowhere.

This is what I've done so far. I split the stereo track and convert both tracks to mono. S represents stereo, L represents left, R represents right, and C represents center. L = L+C, R = R+C, and S = 2C+L+R. The 2C is because both channels have C in common. If I were able to subtract just one of the C's from S, it would be a piece of cake from there, but I can't figure out how to do it. If a version of S would only have one C, I could then invert it and play it with the original 2 mono tracks. This would cancel out L, one C, and R, leaving me with just C, which is what I want. With the way it is now, I always end up with either L+C, R+C, -L+R, or L+-R (- meaning negative or inverted). You can't isolate the center with any of these. I don't see how Voice Trap, Kn0ck0ut, or any of those other programs do it.

[steve]

You're quite correct Thunderbolt, once you've spent long enough going crazy with L+C+R-L+R-C... you come to the inevitable (and correct) conclusion that it can't be done (at least, not that way). The specialist tools use different techniques from this to achieve the impossible. The "secret ingredient" is what they do with FFT analysis.
If you want to go completely crazy, have a read about FFT here: http://en.wikipedia.org/wiki/Fast_Fourier_transform

[Ian T.]

I'm going through old demos that have been transferred from cassette and have a similar problem. What I want to do is lift a female backing vocal from an early version of a song and place it on a later version.

There's not a lot of instrumentation going on at the same time, but there is a lower male vocal and light keyboards and bass. As the female bv is in a higher range than everything else, it seems like I should be able to select the range and eliminate everything else - does anyone know a way to do this? Even a reasonable cleanup (with the male vocal barely audible alongside) would do, but the female vocal does jump out, so it should be possible to isolate it somehow to Paste into the later version of the song,

Cheers,

Ian T.

[steve]

There's a very wide range of frequencies in a voice (select part of a voice recording in Audacity, then select "Plot Spectrum" from the Analyze menu to see a graph of the frequencies). There will be an enormous amount of overlap between the two voices, making it impossible to separate them with frequency filtering.

[Ian T.]

There will be an enormous amount of overlap between the two voices, making it impossible to separate them with frequency filtering.

Thanks, Steve, I figured this might be the case. I have tried overlaying an almost identical piece in Mono without the female backing vocal (a later verse), Inverted and merged, but no success, not sure why,

Cheers,

Ian T.