I have two stereo tracks with the same music but different vocals. What I want is to get the music track without either of the vocal parts. This is proving surprisingly difficult for me. I understand that you can invert one track with respect to the other and get just the two voices(with one being inverted) but this does not seem to help in the process of getting the music.
It seems like there should be an easy way to keep what the two tracks have in common while subtracting the additional elements but the inversion process does not seem to help.
For isolating what is common to two tracks you could try …
Robert J. H. 's stereo tool plug-in … https://forum.audacityteam.org/t/karaoke-rotation-panning-more/30112/1
or Kn0ck0ut plugin [ windows only ] … http://www.freewebs.com/st3pan0va/
Don’t get your hopes up though : isolating the bits you want without any of the bits you don’t is rarely possible.
Thanks Trebor, I will look into these. The thing is that I WAS able to get great results for the vocals only using the usual phase cancellation method. This means that the music was cancelling out perfectly.
What then seems weird to me about my predicament is that there is no way to get what is being cancelled out and the track with crisp vocals isn’t helpful because one of the vocals is inverted. Shouldn’t there be some process for saving what is cancelled out in phase cancellation? Maybe this is naïve but you would think that keeping the similarity between tracks would be even easier than isolating the differences.
It is weird that you are left with the voice only. Normally, it is the other way around.
If you’re using my tool, you can try a +/- 45 degree rotation to put the voice in the center and to remove it then.
Or you isolate the vocals/center which should be the music in your case, in dual mono though.
The problem of isolation is not trivial because information gets lost during phase cancellation.
Imagine the 4 extreme cases:
L - R = D
1 1 = 0
1 -1 = 2
-1 1 = -2
-1 -1 = 0
You see that two states are ambiguous.
My tool (and others) work therefore with 90 degree differences, rather than with 180.
By the way, “Stereo to mono” is the exact opposite of the phase cancellation method, it has mainly the center information but also a lot of side information, just 6 dB less than in the stereo version.
Thanks Robert. I appreciate your help. I am pretty new to this sort of thing and so I certainly don’t understand very much about it. There appears to be a misunderstanding that is probably the result of my using incorrect terminology.
The responses I am getting seem to imply that I am subtracting L from R or some such thing but I am not. I actually have two different complete stereo versions.(I guess I should have been saying 4 tracks) What I am doing is cancelling the music out between the two different versions but I actually want to keep the music without the different vocals. This is why it does make perfect sense that I am getting great results cancelling the music when I line up the tracks perfectly.
I understand what you are saying about information loss and, if I understand the situation correctly, by trying to get the music I am actually trying to save all and only the information which is lost.
Combine the two left tracks into a stereo pair, then apply a plug-in which gives what is common to both.
repeat with the two right tracks , ( but like I said previously don’t get your hopes up).
Ok, all is clear, 2 (stereo) tracks, 4 channels in total.
I suggest that you first download my tool and place it in the Audacity plug-in folder.
After that, you can restart Audacity, select both tracks and use “Remove Vocals”.
With a little bit of luck, one of the two versions will be yet ok.
The nice thing is that my tool returns the music in stereo (if it was a Stereo recording in the first place and not dual mono). You can even apply the effect only on those parts that do only have vocals, thus other sound in the center of the mix is not removed where it isn’t necessary.
You could post a link to the two files (drop box or similar), it is rather complicated to explain it without a real sample.
Anyways, the results can further be improved by rearranging the two center-less tracks.
We should have so far:
Track 1 L R
Track 2 L R
A good time to save the project or the individual tracks (as wav or flac)…
In case that you’re not yet satisfied:
The tracks are hopefully already aligned exactly, otherwise do it now.
You can split those tracks and move the third (mono) track one step up.
Track 1 L
Track 2 L
Track 3 R
Track 4 R
Choose “Make Stereo Track” from the track drop down menu on track 1 and track 3 respectively.
We now want the music that is identical in those L-L/R-R “stereo” tracks.
For this, you have to choose “Isolate Vocals” (or Center) in my tool.
The two tracks have again to be splitted to get 4 tracks. You can immediately make a stereo track (drop down menu again, this time on the second track).
It now looks like this:
Track 1 L
Track 2 L R
Track 3 R
Tracks 1 and 3 are not needed anymore and can be deleted.
That’s just one approach, it really depends on the source material, what you should do.
Hi guys, thanks for your help! I was able to get pretty good results using the “Isolate Center” function and I wanted to report the good news. That is exactly the sort of thing I was looking for and your program did a good job Robert.
I did not do the “remove vocals” suggestion first because there are actually some background/supporting vocals which are common to both tracks which I wanted to keep.
So, what I have at this point is two very natural sounding stereo music tracks(4 channels) with some fainter ghosty vocals at various parts.(in some places it is almost non-existent) In any case, I am very happy with where things are although I am also thinking about fiddleing with it a bit to try to improve it.
Fine that it works so far.
“Remove Vocals” won’t most likely not affect the background vocals, only the lead is removed and whatever is in the center.
By the way, the main difference to “remove Center” is that it takes the filter settings below into account. This means that the low end with the bass and kick isn’t removed. It has some benefit to adjust the low cut to the lead singer’s voice (e.g. about 200 Hz for a female). “Isolate Center” is afterwards appropriate to just keep this low spectrum when used on the LL and RR pairs.
Have you tried to first mix and render the two tracks into one and then apply the “Remove Vocals” effect?
This should theoretically work too, if the two tracks align perfectly and you can reduce the whole procedure enormously.
Thanks for the additional advice. I have it at a point where I am pretty happy with it at the moment. I did try a number of voice removal techniques in addition to “Isolate Center” and I was not happy with the results. It tended to make the music sound unnatural and did not seem to make the voices less noticeable but only more ghostly.
I am not sure I understand the idea about using “Remove Center” since it is the center that I really want to keep. If there were a filter option for one of the “Isolate Center” options then that is something I would consider trying. Still, I didn’t feel like the bass and kick were adversely affected.
I also don’t think they should have been affected. Right? Clearly they would be if you are isolating center by doing L-R but according to our scheme we did it with respect to L-L and R-R and the bass and kick should be common between each of those right?
Yes, I did align the tracks perfectly(at least to the level possible given the sampling or whatever(not really an expert on that))
A = A vocals
B = B vocals
M = music
Just using inversion and nothing else, you have AM and BM. You can invert and get, e.g., B’M’ and AM + B’M’ => AB’. But getting just M by itself is mathematically impossible.