common background separation

Assuming the following situation:

-2 stereo tracks
-each has the same background music/sounds
-one has English spoken words mixed in, other has another language mixed in

is there a way to analyze/process/separate the common background music/sounds based on
those two stereo inputs?


Kn0cK0ut plugin may be worth a try , but don’t get your hopes up.
[ If the voices are simultaneous there will be occasionally be common frequencies within the speech, even though they are in different languages, those sounds will also end up in the “common” result ].

I don’t see how a center-channel remover/isolator is going to help me here. Are you sure you understood my initial question? or maybe I didn’t understand what you meant when you suggested knockout plugin, but last time I checked, I don’t think it can do what I am looking for.

Combine the two left channels as a stereo-pair, then and apply a centre-isolator to that pair, the result will be a mono track, consisting of what is common to the two left channels. Repeat with the two right channels.

You may very well be right : where the two different voices coincide their common-frequency-content will be isolated too.

but the problem is they only share the same background music, HOWEVER, not at the same amplitude… so this trick might not work if it assumes both background music levels will be the same amplitude? if not, i could try it and see what happens…

The discriminating-factor the kn0ck0ut plugin uses is frequency [spectral] , not amplitude.
[ Nor does the phase have to exactly match, as is the case with other isolation methods ].

Kn0ck0ut does include bubbly digital processing artifacts : so the result is not studio-quality.


can I use 0.8 version?

I don’t have Linux for LV2 and not good with compiling code from source. Is there a way to use that newer version under Windows? would be awesome!

I’ve only ever used kn0ck0ut on Windows, just a VST plugin. I’ve never attempted to use kn0ck0ut on Linux.
My copy of kn0ck0ut probably isn’t the latest …
This is the copy of Kn0ck0ut I have (on Windows).png
Again don’t get your hopes up, here is a typical result of Kn0ck0ut : it sounds computery …

Trebor, are you aware that the next version includes the new “Vocal Reduction and Isolation” plug-in?

It is e.g. available in the current RC-3 download of Audacity 2.1.1.

I’ve recorded the beginning of the tutorial you’ve linked to.
The original audio track is only played for a little while and the guy keeps talking over it.
Anyways, the second half is the isolated version (strength 50, normalized).
The results are comparable to knockout, I should think, the more as the source is far from being ideal.


Robert, you might help me out. Im having a hard time with speech and drum separation when the drums are heard in the isolated center. How to get rid of that? Thanks


assuming you are still reading this topic, just one last thing to ask you.

Since I never used this sort of method before and I used a lot of methods, I was wondering about the IDEA behind it.
I mean you mentioned combining two left channels and two right channels and for each apply center-isolation. But why? what is the reasoning behind doing it this way? did you come up with this or did you saw this method somewhere before?