So as many here probably already know, you can invert an audio clip to isolate the differences between 2 audio clips. I’m trying to figure out a way to remove the differences between 2 audio clips, that way you are left only with is the same among both tracks. This is because I have an instrumental of a song that is identical to the version with vocals except for the fact that the instrumental version has an extra guitar melody. I’m trying to remove this guitar track from the instrumental so it isn’t isolated along with the vocals when I use phase cancellation to isolate them. Anyone have any ideas?
It is not possible to isolate the common sound through phase cancellation, but the “Isolate” option in “Vocal Reduction and Isolation” may be able to do something useful. See: http://manual.audacityteam.org/man/vocal_reduction_and_isolation.html
This is because I have an instrumental of a song that is identical to the version with vocals except for the fact that the instrumental version has an extra guitar melody
Are they MP3 sound files? Or iTunes download sound files? These sound files get their tiny efficiency by carefully leaving out some sounds here and there. What they leave out can depend on how many instruments are in the song. Just having a different number of instruments between the two guarantees you can’t use one to cancel out parts of the other.
Vocal Reduction and isolation doesn’t work that way. It can try to use the character of the sound rather than note by note surgical accuracy. We expect it to work better, but it can start to fall apart on some musical notes.
Post back how it went. This is a common request.
Koz
I’m using FLAC files for both. I figured it’s possible to isolate the differences between 2 audio files it would also be possible to remove differences and only differences, or isolate everything that remains the same throughout.
I’m using FLAC files for both.
They’re FLAC when you got them, but FLAC is not a widely used format. Where did they come from?
You can totally try it yourself. Open both songs one above the other. Audacity will try to play them at the same time (and get really loud).
Select the bottom track by clicking just right of the up button. Effect > Invert.
There. The top track and the bottom track are now mirror images of each other except for the parts we’re not expecting to match. Use the Time Shift Tool (two sideways black arrows) to push the bottom track slightly sooner and later to get the best match. Magnify to get finer adjustments.
You will also need to match the volumes by jockying the volume slider slightly up and down.
Rather than getting a theatrically significant cancellation, we expect you to go completely nuthouse trying to get the right match (which is normal).
Warn the neighbors.
Koz
You are bucking a lesser known problem as well. No producer I ever met will produce a vocal without “fine tuning” the original backing track a bit. It’s a medical condition affecting all producers called the DiQuithitt Effect from its discoverer Simon DiQuithitt.
Yet another reason they’re not going to match.
Koz
They were converted to FLAC from the audio files of a video game. I guess I’ll just have to keep playing with it until it sounds right.
It is possible. First, isolate the differences by inverting the tracks like you said. Then, import both of the original tracks as well as that inverted track you just created. This should cancel out the differences, leaving you with only the similarities.
It is possible. First, isolate the differences by inverting the tracks like you said. Then, import both of the original tracks as well as that inverted track you just created. This should cancel out the differences, leaving you with only the similarities.
No… That doesn’t work… I know it seems like it should work but no matter how much you play around with simultaneous equations there is no solution. 
If you have A & B there are 4 “basic” possibilities -
A
B
A+B
A-B
Or, you can get other combinations like 2A-B, etc. but there is no simple math that eliminates the common information. What you want is a [u]union[/u].
The regular-old vocal remover is simply L-R but I assume the Vocal Isolation effect works in the frequency domain (using FFT) and it works imperfectly.