Remove each channel from the other

I’ve recorded an audio interview with two participants, into two separate files. Ideally what would have happened is that my voice was in one file and the interviewee’s was in the other. However, because we were sat quite close together and my lapel mics aren’t all that directional, what we’ve got is both voices in both files – in one, my voice is louder, and in the other his voice is louder. What I’d like to do is, sort of, “subtract” file 1 from file 2, and “subtract” file 2 from file 1. I’ve tried playing around with putting both files in as channels in a stereo track, inverting one, and mixing, but that doesn’t seem to help. I’ve also tried using a Gate, but there’s not enough of a clear loudness difference to Gate away the quieter voice only. Are there other approaches I can try here? (“Get better mics and sit further apart” I know about, and will be doing for the next interview :slight_smile:)

Subtracting one channel from the other requires that they be “in phase”. Because the two mics are in different physical locations there is a slight (probably a few milliseconds) delay from one to the other, worse that delay is going to shift as you moved around, which is going to change the amplitude as well. You might be able to do this for small snippets by carefully adjusting the delay and amplitude of the channel being subtracted.

There are probably ways this can be done using correlation functions and clever software, the problem is akin to echo cancellation, but there is nothing that I am aware of in Audacity to achieve this.

The “gate” function was probably your best bet. Something that monitored the volume of the two channels and muted the softer one might work (assuming you didn’ t both talk at once).

Otherwise I can suggest going through by hand and muting the channels when the wrong voice is talking. (Again doesn’t help when you both talked at once).

Yeah, that’s what I’m doing currently, which mostly works but (as you note) is problematic if we both speak at once, and is exceedingly time-consuming besides. I admit I was vaguely hoping that someone would say “just use the ‘Next Time Sit Further Apart’ LADSPA plugin and it’ll fix it”, but I’m a bit short on sound engineering knowledge. :slight_smile: Thank you for the pointers!

AutoDuck a copy each track with a copy of the other ? : it will attenuate your voice on their recording and vice versa , ( I think :question: ), may cause unwanted attenuation where you are both speaking at the same time.

You can spectrally-subtract one recording from another using plugins like Kn0ck0ut , or RJH stereo tool, however digital artifacts are generated in the process : the result sounds computery.