Can I "subtract" a voice from a audio recording using a recording of just the voice?

My friend and I recorded us playing a game together and we both used two audio channels: our desktop audio, and an audio channel with just our voice.

I wanted to combine the two videos together but it turns out my desktop audio was recording his voice as well (screwed up the virtual audio cable). I was wondering if it was possible to use the recording of just his voice as a way of telling audacity exactly what sounds to “subtract” from my desktop audio recording.

I tried using the vocal remover function but it completely ruins the rest of the audio and doesn’t really get rid of his voice. So can I use his vocals to tell audacity somehow what to exactly remove from the mixed audio so I can just get pure audio without his voice in it?

Thanks in advance.

Yes. Use the kn0ck0ut VST plugin and turn both tracks into mono ones. Pan the video’s audio to the left and the part with only his voice to the right. Make that a stereo track and run kn0ck0ut. It’ll do spectral subtraction and more or less eliminate that voice and keep the other stuff intact.

My guess is that won’t work very well, because it will suffer the same problems as Audacity’s Noise Reduction effect - namely that the voice in one recording is not sufficiently similar to the voice in the other recording.

Post back how well it worked. This is a common problem and it would be good to have a posted solution.

Live recording is a real joy, isn’t it? Orchestra and band recordings run into similar problems.


Under ideal conditions subtraction can work perfectly. Mixing is done by summation so if I have a mix of A+B I can invert A and mix that in giving A+B+(-A) = B.

But usually if you have a perfect copy of A and a “simple pure mix”, you already have a perfect copy of B also.

The regular/traditional vocal remover also works by subtraction and can also work “perfectly” if the vocals are perfectly centered (identical and in-phase in both channels). But, it will perfectly-remove everything in the center (except you can keep the bass and extreme highs) and you end-up with a mono file.

Seems like it deletes the right stereo channel with my friend’s voice in it, and then just makes the left channel (the one with video audio and my friend’s voice) quieter without removing his voice.

This is definitely what I am trying to achieve, it’s just that I am a complete newbie to audio editing and don’t know how to go about this…
My friend and I both recorded desktop audio plus our own voices in a separate audio channel. My desktop audio includes his voice picked up through Discord plus the game I was playing. I would imagine my recording would not have a perfect copy of his voice as the audio quality was surely compressed when he was talking through Discord.