Subtract one track from another..


I have two tracks, one contains just vocals, and the other contains vocals and backing. Is it possible to end up with a track containing just the backing, by subtracting the vocal track from the other one, and how can I do it?


If they’re a linked stereo track, you need to unlink them with Black Down Arrow >Split Stereo Track. Then make them both mono (same dialog box).

Select the voice track only and Effect > Invert. Then select both tracks and Project > Quick Mix.

I’ll put a lot of money that it won’t work. Unless you have access to the original performance master tapes and export tracks, the mix and the vocal will not match and therefore will not cancel out. You can’t just use one singer singing the same song on two different days.


I am trying to do something similar without success…

Someone has taken a song and put additional audio on it. I have a copy of the original song, which is IDENTICAL. I want to subtract this from the modified sound file (or at least reduce it as much as possible) so that I am left only with the additional audio that has been added. I have tried the ‘invert’ method, but this does not work, and just sounds almost exactly the same as the first file, as if nothing at all has been done to it.

So if I’ve got an identical copy of some sound, how do I subtract it away, or at least reduce it, so I can clearly hear what has been added to the modified version?

I tried it using two copies of just the original song and it cancelled out to absolute silence. So then when I add the second version with the additional audio, and get them absolutely perfectly lined up, why doesn’t it cancel out the original song in the same way? What has gone wrong??!?

The only explanation that I can think of is that they are NOT IDENTICAL (though they may be extremely similar). For cancellation to occur the audio samples must be exactly the same value at exactly the same time but with reverse polarity.


You have to use an application that aligns the signals at the millisecond -or rather: sample- level. As you correctly stated, once that is done, the phase/shift between the 2 signals becomes zero and the subtraction is trivial.

I know of at least one application, and am looking for others as we speak (in 2015). To posters johnrichki, Loonwolf and any other forum participants interested: please send an e-mail to ramon at patriot dot net for the name of the free application since I have been sternly warned to “cease and desist” from mentioning any non-Audacity products, in these otherwise outstanding forums.



Audio track may be aligned to sample level using Audacity, which Loonwolf indicated that he did.

That is untrue. I respectfully requested that you “desist from misguiding other forum users”.
This is the post where I wrote that:

Let’s allow the readers be the judges:

In this particular post, a participant (johnrichki) was looking for a specific solution, right? Other posters joined the request.

(a) You provided 0% of what the user needed.

(b) I provided 100% of what the user needed (the “Audio DiffMaker” application)

… and yet I am the one who should keep his mouth shut?


So since it seems that the whole cease and desist thing was just a misunderstanding, can you please post the name of that program now?

I’m having some trouble getting phase cancellation to work. Maybe someone can help me? I’m trying to isolate the piano track of Sympathy for the Devil by phase cancelling the original track with a version that has no piano track in the mix (except for a small amount of bleed from the other tracks). Here is the audacity project for the version with no piano:

The name was already given. And I see no reason to withhold a link What Steve was saying is that app would not help the user we were trying to help.


As has been posted before, cancellation techniques almost never work because the two files do not contain identical works and that can be traced to a pass through compressed formats such as MP3.

MP3 gets its small file sizes by creating cleverly hidden sound damage and the process depends on content. If you have an MP3 track with violins, trumpet and drums and a second MP3 track with the drums taken from the exact same performance, the two drum tracks are different.

The effect is significant. There is a technique in AudioBook production where you intentionally create sound files a bit “off” such that the conversion to MP3 and resulting changes don’t violate publisher’s standards.