Ableton Forum

What's the easiest way to isolate Vocals?

Mcgyyver wrote:What's the easiest way to isolate Vocals?

I've done this just a few times with mixed results. There are many aspects and methods to this but one classic approach is finding an instrumental mix, put that and the vocal on two tracks, beat sync the instrumental to the vocal version so that they beatlock, match the volumes more or less exactly and reverse the polarity in the instrumental (Done with Utility preset "inverse phase" in Live). Then adjust the volume of the instrumental until you get a clean(ish) vocal. Not all Instrumentals, if there even are any, are the same mix just without vocals so results may differ.

Also look at remixes that may be coming in a remix version and a dub version that sometimes are quite similar.

The result will not be perfect likely but it will possibly be easier to mix with your beats/music already at this stage. Then use EQ and envelope filtering, possibly with gating to carve out the rest of the background.

If you don't have an instrumental or remix to use you probably need to look at sections without vocals that seem to sound similar to sections with vocals and painstakingly copy and paste these into place in different tracks. That might be more suitable for shorter bits and not full songs.

A more modern approach would be to use dedicated tools for this task, that I assume are out there, or something like Melodyne that potentially allow more detailed editing than EQs, filters, and gates. It's no walk in the park necessarily, but great fun.

I'm not sure if any of these methods are easy though. If you're lucky or manage to choose the music well they might be.

Hopefully someone that have done this kind of thing more often will chime in now…

What Stromkraft said.
But as your question is very general and sounding like "What's the easiest way to
isolate vocals from ANY song?":

For most songs there is NO way, not even a hard one, to get a usable result.

Imagine a Cake


and now.. how do somebody remove only the sugar from it ?

remove all sugar from the cherrys
remove all sugar from the cream
remove all sugar from the choclate

once it is mixed, its hard to unmix - called entropy
ist easy to produce shit-conterminated water, but even if you
pick out the big pieces you would not wanna drink it.

same is with vocals or any instrument.

with single drum-hits containing a little piece of a pad in backround you maybe more lucky
then with a angelic voicepad containig a harsh guitar-strike.

Well, actually there are ways to only remove the sugar.
But they are not easy and what will remain won't look
as nice as the cake did before

Sometimes the vocals are in the direct centre while all other material is fully panned for stereo and you can (with some fiddly work) just extract the "common" data between L/R then use a little EQ to filter out some of the other business. Sometimes (like on older 70s stuff such as Pink Floyd, Jimi Hendrix etc) the vocals are hard panned to L/R so you can easily snap it. Also if you look for dolby 5.1 mixed film clips (sometimes available on music DVDs) the vocals are mixed directly in centre and easy to get.

As Stromcraft suggested, using instrumental phase cancellation is probably the most common "hack" method. The best method is getting actual acapella versions of the songs.

Sometimes just EQing material either side of the vocals and using some notch filtering to try and reduce some instruments within the vocal range then layering enough of your own sound over the top can work too.

jestermgee wrote:Sometimes the vocals are in the direct centre while all other material is fully panned for stereo and you can (with some fiddly work) just extract the "common" data between L/R then use a little EQ to filter out some of the other business.

The centerness of vocals is indeed something to investigate once you are comfortable getting into the sum and differences of mid and side parts of the stereo signal. Definitely moving into a more tricky if potentially rewarding area though.

One type of material which with I'd like to do something like this is to remove music from spoken word samples. While the music makes it sound more like "a sample", potentially in a good way, getting rid of it opens up many other possibilities for further treatments.