I've seen a lot of tutorials about how to remove vocals and such with audacity, which never seem to work. But I want to know if I can find a song like on YouTube or something and remove all the lead parts, be they vocal or guitar or what ever.

Could you do it say on this song, remove the leads but keep the drums and bass and stuff:

  • 7
    Simple answer, no; for the same reason the 'remove vocals' tricks never actually work either.
    – Tetsujin
    Sep 1 '15 at 7:12
  • 3
    Is it because all the stuff is already mixed together and you can't undo it? Like making a cake batter and trying to just remove the flour, can't do it anymore? And other backing tracks would've just been made by "re-creating" the song but leaving out what ever you want to? Or something along those lines.
    – Cuculoco
    Sep 1 '15 at 10:03
  • 4
    You pretty much have it, on both the cake analogy & how such as karaoke tracks are done. Games like Guitar Hero are done by actually obtaining copies the original multi-track recordings & re-mixing to suit the game.
    – Tetsujin
    Sep 1 '15 at 10:05
  • Added the complete youtube link for you. No reason to have a broken youtube link.
    – Neil Meyer
    Sep 1 '15 at 11:58

Removing vocals or other tracks from an already mixed and mastered file is very complex and depends on a lot of factors. There's no one size fits all answer because tracks can be mixed in all sorts of ways with many different effects.

In the case of Karaoke or Guitar hero, they typically get access to the master tracks and in the case of Karaoke mix it again with the vocals muted, or in guitar hero make several mixes, one for each instrument. The ideal way to make a backing track is to get the masters provided for you (very unlikely, unless you have a good reason and probably money), or look to see if the separate tracks have ever been released in a producer edition (several Dream Theatre and Lamb of God albums have been released with this as a bonus for instance). If the tracks have appeared on a guitar hero game you can usually find copies of the audio that have been pulled from the game.

Trying to remove vocals from an already mixed track:

To do it with a finished and mixed product is much harder. The way the tool works in audacity and others (I'm assuming) is to take the left and right tracks, and then compare to find which parts of the signal are the same (ie, centre mixed and coming from both speakers equally). This is usually your lead vocal, as well as some other things like Kick drum, lead guitar, etc depending on the mix.

With a track that is only the parts that match you can then remove that part of the signal from both tracks to remove that content. If you wanted to be more precise, you could use high and low pass filters to try and filter down to only the vocal content, leaving things like the kick alone. But you'll still lose the content of those sounds that exists in the frequency ranges of the vocals.

An easy way to do this would be to take the left and right track, invert the phase of one, then mix down to Mono. This would cancel out any of the signal which is the same in both speakers and leave only the differences. If vocals are centre panned and dry they would be cancelled out, as would other things.

The reason this doesn't work every time:

There is a big problem with this though: it only works to remove content that is the SAME between speakers. If there's are stereo effects which differ between speakers (reverb, chorus, delay), backing vocals panned to the sides, doubled vocal tracks panned to the sides some, that content will not perfectly cancel and will stay audible. Other mastering tricks like stereo widening can also make this harder to do.


A more practical approach

If you are searching for a temporary solution to, let's say use the backing tracks for improvising over them without being disturbed by vocals or mid-range buzz, there's some little tricks you can do with audio material in a simple sound-editor without the need of sophisticated and expensive tools.

Of course, because of the (in)-fidelity, the result is not usable for production-purposes!!!

  • A multi-band graphic EQ that filters out all the frequencies in the low-mid and mid range. Provide a curve like ---______------------

  • An EQ with a bell-characteristic and a steep curve Q-factor to achieve a similar effect as the graphic EQ

  • Put the same audio track on two tracks of your audio-editor or DAW and put a Lowpass and a Highpass filter on one track to cut the low and high frequencies. What remains is a midrange signal (depending on the settings of your two filters). Now phase invert one of the two audio tracks you've got. This will nullify the mid-range completely...

You can save these settings in your editor of choice and use it as a 'preset' just by assigning a new song to your audio track(s).

All tree methods serve to get rid of the mid range and do in a simple way what more sophisticated tools are doing in a dynamical time and context-sensitive way - but, depending on the material, not always very good. I showed you the hardcore way that will eliminate the stuff for sure, independent of what the source material is.

But you have to keep in mind that both tracks (version 3) have to have the same volume in order to get the midrange 100% silenced. If you're shifting the volume-ratio by changing the volume of either one track you will get back the midrange step by step.

Of course, by cutting the midrange so bluntly you are also killing a lot of instruments that form the chord/harmony portion of the composition. You are essentially left with bass-drum, bass, low-harmonies, HH, Cymbals, high percussions regarding the rhythm-group. And that is what you actually want when you're searching for a play/sing-along track. But anyways you can regulate the effect by changing the cut-off frequency of both of your filters to narrow down the midrange in order not to loose too much.

Hope that that's it what you've been searching for...

EDIT: If you wanna get fancy with method 3 you can also add a third track and phase-reverse one of its channels (L or R) in case your source is stereo. With that - everything that is usually mono like - the 'dry' vocal signal, centered mono effects etc. - will get removed. Only the effect part of each signal which is on both channels like reverberation an the like will remain. Also effects that are using a modulation of the signal's phase like stereo chorus/phlanger/phaser etc. for e.g. pad-sounds. But usually these 'wet' signals are not very loud...

That signal will now fill the hole in the frequency range that you left with method 3 and give you back some of the harmonical context that might be missing after the brick-wall-method 3. But also the 'wet' signal of the things you want to get rid of.

But because of being on its own track you can adjust the volume of your new midrange to your liking.

And of course - the additional track of this extension of method 3 must have the same filters applied to it than the other phase inverted track so that you actually fill the very hole that you dug in the first place. If you didn't do that you would definitely kill all your BD, base and other signals that are typically centered and more or less 'dry'.

But by storing your settings as a preset you have to do this work only once and can apply it later on to any song you like. You may just have to tweak your preset every now and then...

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.