0% Interest for 36 Months! Learn more »
(800) 222-4700
  • Español: (800) 222-4701
Presidents' Day Sale!

Sweetwater Forums [Archived]

After 15 years of great discussions, the Sweetwater Forums are now closed and preserved as a "read-only" resource. For discussions about current gear, check us out on Facebook, YouTube, inSync, and our Knowledge Base.

Can someone show me how to "extract" vocals from a song?


What technique/tecniques should I use if I want to extract the vocals from a song that's been saved as a mp3 file, i e completely silence as much as possible of the instruments and making a track with only the vocals left?
Here is the file I want to extract the vocals from:
May 30, 2010 @06:34pm

This is popping up again for some reason. Extracting or eliminating vocals is not possible. You can EQ out some of the frequencies that the voices reside on. This will also remove anything else sitting on those frequencies. There are several products both software and hardware that claim to remove vocals. However, the results are mostly of unacceptable quality. This has been covered repeatedly on this forum in various places. Check this sticky in the FAQ for more information.
June 1, 2010 @01:55am

First realize what you are asking.....
Here is a stereo waveform.
Please tell us where the 'vocals' are and where all the instruments are.....
There is a voltage curve for the right channel and a voltage curve for the left channel, with the numbers being sampled many thousand of times a second. THERE IS NO SEPARATE INFORMATION.
(computers really aren't magic like they are on TV and movies)
June 2, 2010 @02:21pm

Maybe I should tell you where I got the idea from. I was listening to this song on my iPod one day, when suddenly I accidentally tugged the earphone cord too hard. As a result, suddenly I barely heard the vocals (and that is being optimistic, they were like completely gone), but the guitars and drums (except that the cymbal sound got thinned out) was almost perfect and still crystal clear.
If I could simulate the way my earphones got damaged by completely cutting out the same frequencies/stuff that got lost I should achieve a similar effect, right? Now if I instead could reverse this, and keep the frequencies that got lost in my earphones and remove everything else, then I should have perfectly audible vocals and almost no guitar/drum sound, right?
I found out about a program called Audacity and played around with the equalizers, but I couldn't even get it to be the same "perfectly audible guitar and drums but vocals getting cut out due to certain frequencies lost" effect. In other words, not even close to be able to lift out just the vocals as well.
I based my work around the facts that I know that the vocals are very highpitched and in comparison the guitars are very emphazised in the mid range but, by normal metal production standards, are pretty low in the bass and higher frequencies. So I tried manipulating the equalizers bearing that in mind.
The reason I asked here was in case someone knew a better method/could try and do it for me, since I ran out of options due to my own very limited knowledge/skill with audio manipulation.
Of course, I was not at all expecting a perfect result.
EDIT: If you would like to try your hand the audio file I want to lift out the vocals from is in the first post.
June 2, 2010 @02:34pm

It sounds like the vocals for this particular song are heavily panned to one side or the other, depending on which of your earbuds crapped out. You could just isolate whichever side this is (again depending on what you want to isolate), however, you'll also get whatever else is panned to that side. Vocal extraction isn't really possible, much less vocal isolation. Vocals fluctuate so much that it is impossible to account for all of the changes effectively enough to isolate and remove them.
BTW, I can't figure out how to download the file you've posted. It plays along with all kinds of other spammy ads and crap, but I can't download the file. Try uploading it to a different share site.
June 2, 2010 @03:23pm

I wouldn't worry about vocals fluctuating too much on this one. You'll get what I mean :P
It wasn't one earbud that crapped out, both were affected in the same way (the vocals are centralized). Something in the earphones freaked out and prevented them from playing the vocals anyway :P
Here's the file again:
June 2, 2010 @03:38pm

I don't know what it was that you heard, but there's no way you're going to adequately separate anything in that. Just for fun, I tried Adobe Audition's center channel extractor with all kinds of different settings. There's no way you're pulling anything out of that. It seems like the lead guitar is sitting front and center. It's all basically sitting in the same frequency range too. Not to mention they compressed the crap out of everything so it's just loud. Loud, loud, loud.
Vocal extraction doesn't work unless you have the voice isolated in the mix. Even then, it's witchcraft at best, with more-than-likely shoddy results.
June 2, 2010 @04:06pm

This was the best I could muster at reproducing what my earphones did on my own http://www.wikiupload.com/iKR1GMi1
The difference was that the guitars were less muffled out, and the vocals were even less audible.
What I wanted to achieve was of course the reverse (silent music, audible vocals), but I guess I should give up?
June 2, 2010 @08:21pm

That's what I've been basically saying. You're welcome to keep trying, but you will never achieve the results you are looking for.
June 2, 2010 @10:54pm

That's what I've been basically saying. You're welcome to keep trying, but you will never achieve the results you are looking for.

Actually it is quite simple to achieve the results IF you know that the vocals have been panned center when the engineer was mixing them. Luckily, almost all vocals are panned center.
Anyway, when you import your mp3's into your DAW, make sure you import them as separate left and right mono tracks, so instead of one stereo track, you get two. Now, do it again, importing your mp3 as two separate mono tracks, left and right. Label your four tracks, and make sure to pan them either left and right depending on how you imported them. Now, flip the second two tracks out of phase, and anything that was panned center in the original recording will be phased out.
I literally just did this to confirm the directions I was typing, and all I got was a tiny bit of vocal. Unless the vocal is panned exactly center and things like back-up vocals and what not are not panned somewhere else in the stereo field, the vocals will disappear completely.
Also, you must remember that you are doing sound wave cancelling, so if you solo just one of those tracks, you will hear all of the vocals again. So now those four tracks should be linked together to always achieve the sort of "anti-vocals" sound.
Does this make sense?
April 6, 2015 @02:24pm