blending tracks to extract drums from them. (read the post)

[CastleWorldKing]

Hi Guys!
Just wondering, I have two soundtracks that have different melodies but both have the EXACT same drums.
Is there a way I can remove everything from the two tracks except the parts that are identical, so that I can get just the drums?
Thanks heaps.
P.S: I am using Audacity version 2.0.6 on Windows 8.1

[DVDdoug]

There is no solution.

You can add or subtract the tracks. What you have is two equations:

Track1 = Melody1 + Drums
Track2 = Meoldy2 + Drums

If you subtract those two, you can get Melody1 - Melody2 (with no drums). Or, you can add the two and get Melody1 + Melody2 + 2Drums. But, there's no way to get drums-only. ...If you try to go farther and add or subtract the resulting no-drums file to get drums-only, you'll find yourself going in mathematical circles.





NOTES:
Mixing is done by addition. If you want to subtract, you simply invert one of the tracks mix. (Adding a negative is the same as subtraction.)

It's rather unlikely that the two drum tracks are digitally identical. It would have to be the same drum recording mixed with the melodies at the exact same level with no effects or level adjustments before or after mixing... I'm betting that subtraction does not remove the drums.

If you mix two unrelated tracks, such as Melody1 and Melody2, addition and subtraction sound identical because all of the data is uncorrelated.

If you record yourself saying "Hello" twice, addition will sound identical to subtraction. However, if you take a 2nd copy of the same recording, addition will make it twice as loud and subtraction will give you dead-silence.

[CastleWorldKing]

Could I maybe subtract the tracks to get melody1, then use it on the melody1+drums track to get the drums that way?

[kozikowski]

You want the drum equivalent of Vocal Isolation. As above, vocal removal is easy. Flip one track over and add. Anything the same on both tracks vanishes. It doesn't work the other way. Someone in the past forum threads carried this out to pages of math and never quite got there. "How about if you flip over the second track while you hold your tongue just right......."

Vocal Isolation generates an algorithm for a specific performance that uses gating and processing to pass only the voice. It's rarely perfect, so you can't use it for a performance, but it works far better than the simple math process.

At one time there was a move to include Vocal Isolation with the Vocal Removal tool in Audacity. It may be there in 2.1.1. I'm not in a good place to look.

Koz

[CastleWorldKing]

ok. thanks for your help!

[Robert J. H.]

You want the drum equivalent of Vocal Isolation. As above, vocal removal is easy. Flip one track over and add. Anything the same on both tracks vanishes. It doesn't work the other way. Someone in the past forum threads carried this out to pages of math and never quite got there. "How about if you flip over the second track while you hold your tongue just right......."

Vocal Isolation generates an algorithm for a specific performance that uses gating and processing to pass only the voice. It's rarely perfect, so you can't use it for a performance, but it works far better than the simple math process.

At one time there was a move to include Vocal Isolation with the Vocal Removal tool in Audacity. It may be there in 2.1.1. I'm not in a good place to look.

Koz

Shure enough, It has been added to Audacity 2.1.1.

Thus, you can isolate the drums - at least in theory.
You have to split your two tracks to mono.
Combine the two left and the two right channels into two stereo tracks.
Isolate the center with an arbitrary strength.
Again, split the two tracks to mono and delete the secondary channels (they are the same as the first ones) and combine the two remaining channels to one stereo track.

It is clear that proper alignment is essential.
Drum hits cover most of the frequency range and you will loose the high frequency portion if the alignment is off by more than a few samples.

Any usage of a non-linear phase filter (such as low- and highpass) degrades the group-delays and the phases and the effect won't be satisfying.

You should give us some sample files if you want to have more explanations or how-tos.

Robert