Isolate People in Conversation

[jakubpon]

So I have been trying to figure out this quandry for a little while now and have tried a few things but I haven't been able to get it to work:

I have two audio files from two different Lav mics on people ahving a conversation. The quality on the Lav from person A is a lot higher than the quality on the lav from person B (Technical difficulties).

The sounds from the conversation bleed over between lavs, Person B can be heard on the lav worn by person A, and vica versa.

I would like to silence the recording of Person B whenever Person A is talking and silence the recording of Person A whenever Person B is talking. Is there a non-manual way to do this?

I have tried the center channel extractor and it did have some effect but I was wondering, since I have a good recording of person A would Audition be able to use that to remove their half of the conversation from the track for person B?

Also I have tried reversing the polarity but it had absolutely no effect?

Thanks for your help. I can provide more details and files if necessary

[SteveG_AudioMasters_]

jakubpon  wrote

I would like to silence the recording of Person B whenever Person A is talking and silence the recording of Person A whenever Person B is talking. Is there a non-manual way to do this?

I have tried the center channel extractor and it did have some effect but I was wondering, since I have a good recording of person A would Audition be able to use that to remove their half of the conversation from the track for person B?

You might be able to do this in two passes using the auto-ducker. Normally this is used to duck a music track under speech, and restore it when the speech isn't there, but in principle there's no reason you couldn't do this with another speech track instead of music. But you'd have to do it twice - once for each track, and you'd have to be pretty careful with the settings.

You won't be able to use any subtractive techniques at all, as they rely on an identical signal available on both tracks, and you have nothing even remotely near to that if you used lavs!

My guess is, though, that you are likely to spend as long setting this up and getting it to work as you would spend just editing the two tracks using the volume automation rubber band, especially as the results are likely to be unpredictable at points where they speak across each other, which is usually what happens somewhere in an interview. At that point, using the ducker method, you end up with no signal at all, as both signals will attempt to mute each other - so you'd have to sort that out manually anyway.