Vocal Isolation Project Problem


I’ve been trying to isolate the vocals off a certain song (Fresh Prince of Bel Air, if you must know).

I noticed that there is a certain rhythm segment that repeats itself throughout the track,
so what I wanted to do was sample that segment from a point where there is no lyrics, repeat it through the duration of the original track, then invert it and mix it with the original to cancel out that repeating segment from the song.

Creative and fool-proof, right? I thought so to, but apparently it doesn’t work.
I did exactly as I planned, and aligned the repeating sample to start at it’s original time but when I play the two tracks together it only cancels out the segment on the exact location where I got the sample and then it keeps going as usual.
(If the original is “Beat, Beat, Beat, Beat” it plays as “Beat, ----, Beat, Beat” as the second one is the segment where I got the sample).

And I know there’s no sync issue because as the song returns to normal it’s unnoticeable that two tracks are played at once.

What am I doing wrong? :cry:

It doesn’t work like that. Waveform cancellation only works when the exact lumps and valleys in the blue waveforms match. Even if a trumpet plays the same measures in exactly the same rhythm, the vibration of the lips on the mouthpiece is going to be different. That’s where the cancellation comes from.

It’s like everybody is playing “G”, so they should all cancel out. No. Not happening. Even if the same artist is playing G on the same instrument it may not cancel out. Breath control vibratto if no other good reason. They’re slightly different notes every time he blows.


I understand that, but I thought this was a special case because it sounds like it was digitally looped throughout the song.
Plus, I’m a noob, but it seemed to me like the wave patterns were the same on the smallest levels, as to suggest that this is in fact correct.

And, in one of my tries I apparently misaligned the sample I took, but it still canceled out some of the sound and made kind of an echo sound, but still only on the part the sample was taken from.

The important thing here is that to get cancellation, the waveforms must be identical. Even the smallest variation will stop the cancellation from fully working. If, for example, they have looped the sound and then applied reverb to the track, the loops will no longer be identical. They also need to be lined up with sample accuracy.

This cancellation method is a way of testing to see if two audio clips are identical.

thanks for the help.

I find it hard, though, to understand why there isn’t a way for me to cancel out a range of sound like I would with a color in photoshop.

The audio equivalent of that would be a band stop filter. You can create that effect by using low-pass and high-pass on two copies of the track. However, that is not the effect that you are describing. The photoshop equivalent of what you are describing would be more like a photoshop filter that could automatically filter out “trees” from a landscape photograph.

<<<The photoshop equivalent of what you are describing would be more like a photoshop filter that could automatically filter out “trees” from a landscape photograph.>>>

Can I get one that filters out bad composition?

Musical performances are rarely one tone. This is an analysis of one single note on a Grand Piano. “G.”


Each of those spikes is a separate musical tone and when you add them all up, it presents as a piano. The richer and fuller the instrument, the more of those tones there are. Isolated flutes tend to be much more like you’re talking about; one primary tone followed by other stuff much higher representing breath sounds. Even at that, it would be hard to manage one flute and not touch anything else.