Vocal isolation/extraction

So. Vocal extraction. Done properly. How in the hell is it done? ¨

For a long time I've tried the usual thing with EQuing the fuck out of the track, soloing the mid, trying to filter out the rest, yet the background instrumentation is still there, not so well-hidden.

How do musicians such as Burial, James Blake, etc., do this so well?

Example: James Blake's 'CMYK' samples Kelis - 'Caught out There' (sample appears at 1.17).

http://www.youtube.com/watch?v=LA-lOywYGic

-

http://www.youtube.com/watch?v=N3JFwd1bk4Q

In Blake's track, there's pretty much no trace left of the instrumentation, and this is quite prominent in the actual original track. Is it, in this case, just a use of re-recording of it perhaps?

Another example is Burial. He can get away with much more considering that his soundscapes are much more sonic and nosier, so if there's any instrumentation left, they'll probably be layered deep under other layers.

Another thing, when talking Burial, is that he claims to use/have been using Sound Forge and several forum discussions about this very subject claims that Sound Forge, somehow, lets you get immense EQ control, so that SF would be the best piece of software to do this in.

Has anybody any idea? Besides "just fuck around with the EQ, mate."

Melodyne perhaps?

http://www.celemony.com/cms/index.php?id=products_editor

Or just acapellas found around the web.

EQ, and i mean proper EQ.. spectrum analyzing helps.. what you should go for is isolating the frequencies of the vocals, not just ruining the audio

chorus FX tend to blur percussion while leaving vocals moderately intact

some aggressive gating to the level of the vocals

reverb that's eq'd to the tone of the vocals

vocoding? etc.. just experiment.

basically, you're going to have to sacrifice some fidelity to destroy the sound in a way that affects the vocals the least.. some mid range instrumentation will usually be left in there and you'll just have to deal with it.

both examples you posted likely used acapellas. but maybe what james blake did was drown the audio in reverb very quickly so he could cut the percussion parts of the sample out + heavy delay.. there are also programs with pitch shifting algorithms that favor tones/vocals before rhythmic content (ableton live for instance), so when you pitch it way up the percussion loses alot of volume/clarity. kelis track is sparsely instrumented as well.. if you listen to the reverb in blake's track some of that could be drum hits

you have the Audacity to do this sort of thing?

:sup:

hi there..

a good read would be..

http://www.acapellas4u.co.uk/topic6806.html

if you are dealing with sequenced stuff, than the PHASE CANCELLATION METHOD is pretty much your best bet..

there are also few VSTs for this kind of job..

http://www.freewebs.com/st3pan0va/

http://www.paulrharvey.co.uk/elevayta/azuifgeh.htm

and the almighty spectral sound processing via fft might also help..for instance

http://www.birds.cornell.edu/brp/raven/RavenOverview.html

cheers,

io

.. forgot about those beauties

old and buggy, but rather wonderful

http://www.freewebs.com/st3pan0va/sp3ctr3.htm

and this monster

http://rekkerd.org/dtblkfx/

cheer.io