Extract 3 tracks from a 2 track recording?

[Arlin]

Just thinking out loud here:
For those on a budget, would it be possible to use a 2 track (stereo) recording and extract 3 separate audio tracks with some clever audio editing?

Here is what I am thinking:
Track #1 - Stuff panned HARD LEFT
Track #2 - Stuff panned CENTER
Track #3 - Stuff panned HARD RIGHT

Is it (mathematically) possible to manipulate the above stereo recording and "filter" out the 3 individual tracks? I imagine using something like audacity to invert and merge the LR tracks in various ways to separate the 3 tracks.

Here is a scenario:
Vocal #1 - panned HARD LEFT
Tracks - panned CENTER
Vocal #2 - panned HARD RIGHT

This just popped into my head. I have not really tried to think it through...

[Gracetech]

In a perfect world this would be possible but there will be issues with this process and it won't in this world be perfect. That said you would usually have a little bleed over anyway acoustically so it's completely possible knowing that it won't be perfect. Now you have to look at how to break down the channels which would be alot of phase flipping and overlaying to isolate each section. I honestly don't think the work is worth while for the minimal gain but if it is worth it for you i could work on a process that would get you what you want.

The process would look something like this.

Apply the left side against the right side with the right out of phase 180 degrees. At this point you can output a new audio file and then import it back in and flip the phase of the right side 180 degrees so it's back to normal and now you have your left and right sides and all you need to do is pan hard right and left to get those singled out. Now import the original track and put the Left/Right track out of phase and you will get your center channel by it's self.

crt

[[Director] Joseph B]

This sounds like a project for someone who has too much time on their hands. I've got a headache just thinking about trying to do it.

[tedanderson]

What if you filtered it through a surround sound system? The one that I have will create a center channel (by mixing the L&R channels) even if the audio is only 2-track. I don't know if there is any additional processing taking place but I can definitively tell the difference between the 3.

[rdj561]

In Cool Edit / Adobe Audition, you can seperate left and right channels, and then create a third channel by mixing the other two. But this is really accomplishing nothing because the resultant audio would sound exactly the same once played through stereo.

Once "mixed" I'm not sure anything can be unmixed accurately. You see them isolating sounds in the movies, but in my experience, it's near impossible. If the sounds are in separate dB and pitch ranges, you might be able to eliminate treble, leaving more bass - or visa versa. But I doubt this would be very usable.

It's like taking an image flattened in Photoshop and then unflattening it to it's original layers... I don't think it's really possible.

[waynehoskins]

Theoretically possible. What comes to mind is (L+R), (L-R) and (R-L). That's easy to mathematically extrapolate, though it would sure be cumbersome to generate. It's the FM Stereo method, except backwards and to another degree -- that system uses mono (L+R) as the main signal, and modulates the L-R difference signal on the stereo subcarrier, relatively low bandwidth.

Much easier, and much more discrete, to get a multitrack interface.

Hard left and right (and center) would be possible too -- I think this is what Chad said. Subtract R from L to get hard-left; L from R for hard-right. Subtract that stereo waveform from the original , and then sum them, for center channel. I think that would work. I've used the first part of that technique to remove the center channel (where vocal normally is). That would be relatively easy. But much easier to be able to take in more than two tracks from the start.

Also note that any distortion (this would be things like MP3 compression techniques) will mess it up because we are interested not in preserving a reasonable listenable facsimile of the original stereo recording, but in doing relatively precise calculations between the two tracks.

[rdj561]

Originally Posted by waynehoskins

Hard left and right (and center) would be possible too -- I think this is what Chad said. Subtract R from L to get hard-left; L from R for hard-right. Subtract that stereo waveform from the original , and then sum them, for center channel.

That's very easy to generate, but when played through normal stereo channels, will not sound any different than LR stereo.

I do this sort of thing every week when adding a shotgun mic's audio (ambient congregation noise - laughs clapping, etc) to our church's house audio.

I'm just worried that the end result wouldn't sound any different than the original.

[Pastor Ryan]

I don't think you can get an entirely different 3rd track.

Being in the middle has some R and some L....

My guess would be if you put a track in the middle it would be some on the right and some on the left.