Voice Isolation from a Mix

[Brett - K A L I S I A]

Guys,

Just wondering, did you have any chance with any tool to extract (isolate) a voice from a mix (I mean a stereo audio file) ?

Cheers,

Brett

 

[Razorjack]

Hmmm...you would need to work within the stereo field (I use SoundHack)...you can mostly remove the vocals from a mix by removing the center of the stereo field so theoretically you could isolate the middle of the stereo field to remove most of the elements of the mix. This won't be perfect though, as in most cases the snare and kick will also be in the center of the field, this can mostly be fixed with filtering though.

 

[black sugar]

How isolated does the voice need to be? Like Razor was saying, that's a pretty tall order if there are drums in the mix.

Is there a Blind Source Separation VST plug yet? Heh.

http://www.ultimatemetal.com/forum/showthread.php?t=213911


edit: Oh, and most "vocal removers" for karaoke and such just flip the phase on one of the stereo channels, which kills the center of the image. I don't know of a way to kill the far L-R of the image.

 

[Matt Crooks]

Well the Dangerous Music S & M can do it... but it is a $1500 hardware box...

 

[~BURNY~]

You can try this one:
http://www.freewebs.com/st3pan0va/

 

[Gomez]

It's called M+S (middle and side) Decoding... a technique nearly as old stereo recording.

There are lots of plugs, hardware that you can do it with. You can do it with any sequncer routing the track the right way...

One of the reasons why fairchild 770 stereo comps way over £10000 these days... they feature M+S decoding. Yum. One day....

oh yeah and it only works if the vocals you want to drop in level are panned in the center.

 

[~BURNY~]

Another one :
http://www.elevayta.com/product13.htm

 

[Brett - K A L I S I A]

Hey thanks,

Yeah I've tried the Elevayta thing, not great, it turned out that the one working best on the kind of material I wanted to work with was Kn0ck0ut, though it's far from being perfect of course

 

[Asmus_Thomsen]

I'll usually do like this

- split the song in L-R, and then copy these to six tracks (3 L and 3 right). 1 pair for the low end, 1 for the middle and one pair for the high end
- Flip the phase on on of the center channels, and don't pan them
- Pan the two lowend tracks hard L-R, same with the two highend tracks
- Low pass and highpass on the lowend and highend, low and high enough to remove as much of the vocal as you can. PErhaps try to isolate the voice in the middlechannels, but make sure you do exact the same thing on the two-centertracks (or else your phase shift may fuck up)

Then just adjust a little here and there. By this you still get a low end (which often disappears), and a nice high end without to much vocal.