Isolating Vocals from Background Music (Part 1)

callas.jpg

“How do you separate the vocals out from the rest of the music?” This question is asked a lot in remixing circles. It is asked a lot because the short answer is so unsatisfactory. The short answer is “you can’t.”

The long answer is longer than Ramadan. This is the first of several entries that will be devoted to answering just that question, hopefully, in a fashion in between. This series will get to the specific tools and techniques used to isolate vocals including ‘before’ and ‘after’ demonstration sound clips. There are, however, important things to get out of the way first: you may want to isolate the vocals from your favorite industrial metal opus but that doesn’t mean the track itself is suitable. This series will spent a fair amount of ink on what kind of source material to look out for, including what to avoid.


While these articles will focus on singers and vocals the same techniques apply to any solo instrument you wish to isolate from the rest of the track such as saxophone, guitar, etc.

The Expectation Bar

You will never actually extract vocals from music that has been mixed together. The best you can hope for is to isolate the vocals so that the original background music is significantly filtered and then literally drown it out with other things in your remix. (Real recording geeks and audiophiles hate the term ‘drown out.’ It’s much cooler to say ‘cancel out’ but to the rest of us the results are the same: you can’t hear it anymore.) In addition there is a very good chance the resulting vocals will have a ‘processed’ sound to them. Just how processed depends on all the factors this series touches on. Sometimes it will sound good enough, but sometimes it will sound as if Madonna is singing through an answer machine — oh, wait, she already sounds like that — well, you know what I mean.

Having said all that, you may be surprised at how good the results can sound. If you think about it, none of these obstacles have stopped thousands of producers and wanna-be producers from trying.

But do you have the right equipment and software? Isn’t there some magic box the heavy producers use?

It is easy to convince yourself of the existence of some magic-voo-doo-box-of-sprinkles that only DJ Food has access to. It is easy to assume that if you paid enough money to the hardware gods you might gain the keys to this kingdom and only then could you sound just like all your nuanced virtuoso turntable heroes. In my make-it-moronic version of this fantasy, some rich fat record company with lots of cash for just such a thing would pay for the hyper-expensive magic box and never actually ask me to repay them for the expense because they trusted me artistically to “do the right thing” with their money.

Meanwhile back on Earth there are only so many ways to manipulate an audio signal. If you have a sound wave editor or a host mixer application like ACID, Sonar, Cubase, etc. you are 90% of the way to having the same tools the “big boys” use. Of course your tools cost a fraction (or were free) and theirs may or may not work a whole lot better. Otherwise, you know, it’s the same thing.

Required But Only Vaguely Illustrative Anecdote

My old Gerrard turntable (similar one here that sold for $1.99 USD the other day on eBay) had its cartridge wires exposed as a matter of ‘techie’ design even though the terms techie, geek and nerd were not in use in those days. There came a time when I had to replace the cartridge and while putting it back I managed to get my wires crossed (no snickering). The next time I played my favorite Jeff Beck album, Beck-Ola (or was it Truth?) I was stunned to hear Beck’s lead guitar even further out in front of the rest of the band than usual. In fact you couldn’t even hear most of the band! I started playing every record I could find and almost each one (but not all) gave me the lead guitar isolated from the rest of the music.

Once the word got out about my magic turntable, my house (or more accurately, my room in my parent’s house) couldn’t have drawn more people if the image of Mother Theresa had emerged from the cracks in my wall after a 6.5 earthquake. Every 17 year old acne ridden minimum wage drop out with a Strat starting streaming in with their favorite vinyl looking to hear the starkly isolated guitar solos. I guess they thought hearing their favorite record this way would help them learn the nuances of their icon’s virtuoso performances. Here comes Highway Star, Tie Your Mother Down, D’yer Maker, Bellbottom Blues

“Dude, your record player is touched or something!”

I never found out exactly how any of that happened but I gather after decades of playing around with audio software that by plugging the cartridge wires onto the wrong pins, I must have reversed the phase polarity of some of the signal therefore canceling some of the other signal out of existence blah da blah blah-blah zzzzzzzz….

That whole episode taught me a very valuable lesson that applies more today than at any other time in my life: there isn’t enough Ozium in the world to withstand another apparition on a turntable in my life. Eventually I put the wires back into place so I could enjoy Quadrophenia at full volume and the rest of my life in peace.

Not Very Transitive Addition

When I started remixing I remembered the “Miracle of St. Gerrard” and hoped it would be illustrative beyond buying stock in air fresheners. The only thing it actually ended up illustrating was what wouldn’t work. Once again, there is no magic bullet for isolating vocals from the background music.

What confuses some folks (ok, me, it confuses me) is how easy it is to remove the vocals from a track. There are very good free plug-ins (like AnalogX’s Vocal Remover) but even better: any channel converter tool will do the same thing with the right settings. Here’s the one built into Sony Sound Forge set to the preset called “Vocal Cut.” Even if you don’t have Sound Forge you should have some kind of “channel converter” tool and you can set the parameters to the same values. The result will be the same, there’s no fancy proprietary algorithm at work here, just addition.

These vocal removers are pretty damn good. Here is an mp3 clip of one bar of music with the vocals, one second of silence and then the same exact bar of music with the Vocal Cut preset assigned to it. So if that’s so easy, where is the “Vocal Isolate” preset?

The vocal removers take advantage of a pretty dependable trait of modern stereophonic recordings, namely that producers and recording engineers will place every single instrument in a mix at least slightly off center except for vocals, bass and bass drum. That means the exact same signal for the vocals is coming out of both right and left channels. Using the same principle at work in my miracle Gerrard turntable, adding a signal to the inversion of itself with result in no sound. Since the vocals are in both speakers exactly the way, if you reverse the left channel and add it into the right channel you cancel the vocals out of that right channel. Because nothing else matches exactly, everything else stays there (or more likely sounds phase shifted.) Do the same in reverse for the left channel and poof! the vocals are gone. Again, this only works because the vocals are sitting 100% in the middle of the mix 99% of the time. If they are skewed by panning or (more likely) panned delay then even this trick won’t work. Of course this technique is also likely to take the bass and bass drum with the vocals into bit-dust-land but it’s a whole lot easier to recreate or reorchestrate a bassline or bass drum pattern than the vocals.

Just in case the logic train hasn’t left you standing at the station, you might have noticed that it should be mathematically possible to do a reverse procedure (isolate the backing using the above technique, then apply the result of that to the original) that leaves just the vocals. There are plenty of articles around that claims it mostly works just fine, but this has not been my experience. I have not been able to isolate vocals good enough for use in a remix using the ‘automated’ techniques implied here. (Readers with different results are encouraged to leave comments.)

So much for irrelevant stories about my childhood and simple steps for doing the opposite of what we came here to do. Next time we cover what to look for in source material.

([UPDATE] Make sure to see TimG’s tips on inversion + noise reduction here.)

Read Part 2 here…

7 thoughts on “Isolating Vocals from Background Music (Part 1)

  1. Jaded

    I’ve read similar tutorials. What they always neglect to mention is that the instrumental mix must be IDENTICAL to the original mix. The other thing that often stops this from working is the difficulty encountered when trying to get both waveforms EXACTLY lined up.

  2. Pingback: beatmixed

  3. Pingback: Brokenlooppers

  4. Camilo Martin

    I do song – instrumental = vocal with kn0k0ut and use it for remixes quite decently. It’s not so bad, even with hard tracks, if you take the time to fix small artifacts and hide the mess with more things on top of it after remixing.
    Please note that I’m not a professional, I’m a normal geek that likes to do that stuff with few favourite songs. It’s fookeeng easy with most popular songs.

  5. Nasser

    soon i will start working on a software that i wish will work, the goal is to isolate the musical instruments that are playing in a song from each other … do u think its doable ?? can it happen ?? .. please help !!!

Comments are closed.