Lets say we have two tracks A and B. Both tracks contain the exact same audio information PLUS uncorrelated noise.
By inverting one of the tracks and mixing both tracks together for subtraction I get the whole noise from track A and track B isolated.
Having the noise from both tracks A & B isolated in a new track looks promising but since the noise wavefrom either form track A or track B is now inverted it seems like a dead end when working with waveforms. Right?
The only thing I can think of is some kind of spectral subtraction. i.e.
Combine track A and B (contains audio information PLUS noise of both tracks)
Subtract isolated noise (noise of both tracks) spectral wise.
I mean even if the noise information of one of the tracks is inverted in the waveform, the spectral reproduction of the isolated noise should be the accurate.
Is there a plugin, method or tool to try this out?
I experimented with both plugins and both are doing exactly that (some sort of). The noise can indeed be reduced.
Main problem of both plugins is, that their resolution is set fairly low and as a result the noise becomes quite blubbery. I don’t understand why there is no option/setting to change the resolution, bands etc. I wouldn’t care to wait 10 minutes or even a hour to get some useful result.
By shifting the frequency range of the audio track to here and there it is possible to get a more natural noise but then than the vocals become blubbery instead. I assume therefore that the plugins are not resolving all frequencies evenly.
By the way… The kn0ck0ut link isn’t working anymore and I couldn’t find v0.9. By visiting webarchive.org I could download at least v0.8.
The “Noise Reduction” can*t work because it applies the “noise profile” blindly over the the whole audio track and has no time differentiation. That works well on constant noises like a single hum or wrestle for example.