Remove Small Section of Center Channel from Stereo Track

[EmpireStrikesBack198]

I was wondering if there was any way to select a small portion of on audio file (while viewing its spectral frequency) and apply a center channel remover to remove the center channel from only that portion? Can this be done with Sound Forge? I prefer to use iZotope Vocal Remover. Below is a picture of what I am speaking about. I only want to remove the center channel from the selection.



Also, my Adobe Audition 3.0 isn't displaying any of the center channel extraction tools. "Vocal Remove" is listed under "Favorites" but it is grayed out. I reinstalled the program, but the tools still aren't there. Anyone know why? Thank you!

[aedipuss]

you might try exporting as 6 channels of wavs and then working on the center channel.

[EmpireStrikesBack198]

you might try exporting as 6 channels of wavs and then working on the center channel.

I am working with a stereo track. I want to remove a portion of the center from the stereo track.

[aedipuss]

as in ac3 2ch.? there is no center channel. it's only "made up" during decoding. you'd need the original source to change it.

[EmpireStrikesBack198]

as in ac3 2ch.? there is no center channel. it's only "made up" during decoding. you'd need the original source to change it.

No, I am working with a WAV. I am working with an original source, which is a stereo file. The music is in the left and right channels and the vocal is in the center channel.

[Noahtuck]

In stereo, IE: 2.0, there is NO center channel......

There are 2 channels, LEFT & RIGHT, 2.0, stereo,

[EmpireStrikesBack198]

In stereo, IE: 2.0, there is NO center channel......

There are 2 channels, LEFT & RIGHT, 2.0, stereo,

Then where are the vocals in a song, then? Maybe I have my terms mixed up...

[aedipuss]

i get what he's after. it's the sound that "appears" center stage when 2 stereo channels are played. it's sound that is the same in both channels at the same time. it's mathematically possible to isolate the same sounds from both channels with something like l-r=d and then (l+r)-2d=c where d is the difference and c is the same sounds or center. i don't know if sound editors can do that for you.

[netmask56]

The algebra of a stereo 2 channels recording with a phantom centre as follow.

Lt (left total) = L +½C
where L = left information
and C = the vocal desired to appear in the centre ie between the speakers

Rt (left total) = R +½C
where R = right information
and C = the vocal desired to appear in the centre ie between the speakers

So if we sum them we get L+R+C ie a mono rendition
or
Subtract one side from the other and we get

L-R ie no vocal (this L-R rendition is also mono) and it is aurally identical to the reverse ie R-L

There is no way one can extract the C component in absolute isolation - do the maths....

So if the OP simply wants to remove the centre component from the entire file this is possible - use vocal remover.
If the OP wants to remove the centre component from part of the song say from one minute in to the two minute point then that can be done simple by marking the area as a selection and applying the filter.

[aedipuss]

you missed the final steps. after you've subtracted the l-r and gotten rid of the center you can then subtract that from the original and leave the center.

[Gavino]

you missed the final steps. after you've subtracted the l-r and gotten rid of the center you can then subtract that from the original and leave the center.

No, you can't.
The 'centre-removed' track you get from the first step is mono.

See discussion here and following.

[AlanHK]

Then where are the vocals in a song, then? Maybe I have my terms mixed up...

So what you actually want to do is remove the vocals?


No guarantees. But see here for methods using Audacity.

[aedipuss]

yes and the step of adding (l+r) is mono also, so you can subtract mono from mono. i'm not saying it's perfect but it might do what is requested.

there aren't 3 things. there is only l and r. no c to start with. c is an artificial construct of the brain when listening to 2 stereo sounds from the left and right and it perceives a tone that is equal in volume from both channels and arriving at exactly the same time as sonically in the middle.

[Gavino]

yes and the step of adding (l+r) is mono also, so you can subtract mono from mono. i'm not saying it's perfect but it might do what is requested.

Did you read the thread I linked?
If L = A + C and R= B +C, you can eliminate C by taking L-R or R-L, but there is no combination of L and R which will give you C on its own. Try coming up with such a combination and you will find it is impossible.

there aren't 3 things. there is only l and r. no c to start with. c is an artificial construct of the brain when listening to 2 stereo sounds from the left and right and it perceives a tone that is equal in volume from both channels and arriving at exactly the same time as sonically in the middle.

You only have L and R to work with (which is basically the root of the problem), but each consists of a number of source signals which for our purposes we can simplify to three:
(1) sounds present in the left channel only;
(2) sounds present in the right channel only;
(3) sounds present in both channels equally.
Given only L and R, you can eliminate (3) as we have described, but there is no way to eliminate both (1) and (2) to leave (3) on its own.

[Cornucopia]

aedipuss, you're cutting corners and then making false presumptions.

That's why in High School, they have you show your intermediate work.

Here's an example in which we have known Lt and Rt channels, and unknown true L and R and C signals:

Call Mono "M"
M = L + R
Mt = Lt + Rt
 
Call Difference "D" (sometimes called "ambience" or "surround")
D = L - R, conversely D = R - L (same signal, but just phase inverted)
Dt = Lt - Rt
 
If :
Lt = L + 1/2C, and
Rt = R + 1/2C
 
Then:
Mt = (L + 1/2C) + (R + 1/2C), or
Mt = L + 1/2C + R + 1/2C, or
Mt = L + R + (1/2C + 1/2C), or
Mt = L + R + C
 
And:
Dt = (L + 1/2C) - (R + 1/2C), or
Dt = L + 1/2C - R - 1/2C, or
Dt = L - R + (1/2C - 1/2C), or
Dt = L - R + 0, or L-R
 
So if you add, you get:
X = Mt + Dt, or
X = (L + R + C) + (L - R), or
X = L + R + C + L - R, or
X = (L + L) + C + (R - R), or
X = 2L + C + 0, or
X = 2L + C, which is twice as loud as
X = L + 1/2C, thus
X = Lt, so
Lt = Mt + Dt
 
If you subtract, you get:
X = Mt - Dt, or
X = (L + R + C) - (L - R), or
X = L + R + C - L + R, or
X = (L - L) + C + (R + R), or
X = 0 + C + 2R, or
X = 2R + C, which is twice as loud as
X = R + 1/2C, thus
X = Rt, so
Rt = Mt - Dt

NOWHERE do you get C alone, or L alone, or R alone. They just "circle back around each other" like binary stars, or opposing warriors.

In that example, and in real life, you can only "work" with what you know (aka the items with "t").

Unlike what you said, there ARE 3 things. L and R and C (without the t) are ALL "constructs" and very real (in a virtual sort of way). C is just coming from a "phantom" speaker.

Like Gavino & Netmask56 said, you can't ever isolate these elements (as they're already interwoven, and in a sense, dependent upon each other). Just like you can't un-mix a soup or un-bake a cake.

Scott