Knock Out

[boon23]

Admin note: This thread originated in the discussion of a FanEdit release. As it relates to a general audio editing issue, it has been moved to a more appropriate location.



Knockout:
Where the vocals are not centre panned you could try a recent free
VST plugin (Windows only) called knOckOut
http://www.freewebs.com/st3pan0va/

which can also be run in Audacity. This works on two mono tracks (split
from a stereo track and made mono as above) by extracting the right
channel frequencies from the left channel frequencies, leaving the "result"
(the frequencies in left channel but not in right) in the former left
channel.

 

[TV's Frink]

Knockout:
Where the vocals are not centre panned

Can you explain what this means?

 

[boon23]

most vocals are usually on the center channel in both a surround track and a stereo track. Center channel extraction is one of the easiest ways of voice removals (if it works). Now Knockout for audacity uses a different method of taking what is on the left (or right channel) and delete it from the right (left) channel. This is done through inversion and can also be done manually.
Each sound is a wave at a specific frequency. With digital sound there is an exact opposite to that wave (inverted) which nullifies it. It's just math.

 

[TV's Frink]

Sorry to keep taking this off topic but:
1) Are you saying stereo tracks have center channels? I can't get my head around that.
2) I understand that if you can find the frequency of the vocal you can nullify it. But how does the left channel relate to the right channel in this case?

 

[boon23]

yes, stereo tracks do have a center channel (and I did not know that either until my last edit).
With Knockout the difference between the left and right channel as mono tracks is separated, which is a different method than using the center channel.
Both methods will in most cases NOT result in getting a clean voice track, but where one does not work the other might work.

 

[ThrowgnCpr]

"stereo" tracks should not have a center channel. Thus, the labeling of "stereo". If it is a 2.0 stereo track, it should just contain left and right tracks. However, if the so called stereo track is a 2.1 track, then it would have a center track. I thought most "stereo" DVD tracks were 2.0, but apparently you guys are saying that many of them are 2.1?

 

[TV's Frink]

You guys? I'm asking the questions here!

 

[herowoac]

This discussion seems a bit of topic, but anyway:

Although I only have very little experience with mastering for cinema,
I do have more experience with stereo mastering for music, and it's quite simple really.

In music: left + right = mono (or center)

In cinema instead they often used so called XY mastering, in which part of the signal in the left and right channel are out of phase with each other, in order to retain more of the complete sound-stage within a single channel. (Theoretically without losing any of the soundstage in the stereo mix) This is done primarily so that the left channel can be used alone without the right channel.

Dolby Pro-Logic used the same basic principles to put a surround soundtrack into a stereo mix.

Hence in standard (music) mastering it's quite easy to remove a voice in the center,
you simply combine left and right, invert the result and combine the inverted signal with the left and right,
and there you go, a center-less stereo mix.
(This is akin to how dolby pro-logic derives it's surround channels by the way, not exactly the same but almost)

With XY-mastering it becomes more complicated, because in an XY-mix the left and right
partially cancel each other out, and hence it becomes much more complicated to derive a true center channel
(It's something like a five step process, certainly can be done but it's a pain to get it right manually)
Hence the need for special tools like knock-out.

By the way I'd love to watch this edit.
PM me the links whenever it's convenient for you.

 

[TV's Frink]

For anyone who missed it, this discussion was moved to its own thread. So feel free to continue talking about it!

 

[Remixed by Jorge]

yes, stereo tracks do have a center channel (and I did not know that either until my last edit).
With Knockout the difference between the left and right channel as mono tracks is separated, which is a different method than using the center channel.
Both methods will in most cases NOT result in getting a clean voice track, but where one does not work the other might work.

When you say channel do you mean tracks? I always thought stereo tracks as L + R.
But if you mean that the combination of L+R create a spatial center audio experience, that i understand.
Like combining red and blue to get magenta.

 

[TMBTM]

I gave up with this invert method...
My goal is usualy to keep the vocals only but that is not possible: the frequencies to invert always are different and so the process is fucked.
To remove the vocals on the center audio? Well, I remove the center audio! If there are still "ghost" vocals on the right and leght, I remove them also and put the movie score in place (and add sound effects when I'm not on a lazy mood).

 

[Gatos]

My goal is usualy to keep the vocals

Me too

 

[ThrowgnCpr]

When you say channel do you mean tracks? I always thought stereo tracks as L + R.
But if you mean that the combination of L+R create a spatial center audio experience, that i understand.
Like combining red and blue to get magenta.

This is how I would have interpreted it too. By boons and herowoac's descriptions it sounds like there are 3 tracks, which makes the stereo name meaningless.

 

[TV's Frink]

Glad I'm not the only one confused...:-o

 

[boon23]

audacity and audition (and probably every good audio editing tool) have a center channel extraction filter. And for that a stereo track is needed (yes, that is 2.0). The center channel is what comes equally from both channels, so it extracts what is mono. This is very often true for the voice track and for sfx, so with this method you can filter out vocals (or at least a lot of them) and sound effects.
Whatever you want to remove from an audio track, this is no exact science. If voice and sound effects and music are mixed into each other, you simply cannot separate them 100%.
But you can get close with various filters.
Here is the center channel extraction method explained to GET the voice track:
1. load the stereo audio track into your audio editing tool.
2. use the filter "center channel extraction" (adobe audition comes with a few more filter goodies like setting the frequency for voices, which can be very helpful)
3. mark the result
4. use the filter inversion to do a 180° flip of the wave structure.
5. copy the sound file.
6. by using the "reverse" function go back to your original unharmed sound file.
7. mix the copied track into the original sound file (at 100% volume).
8. the result should already show clearly that parts have been removed from the original fie. Now listen to it and if this got you any closer to clear voices without music, then you can continue working from there.

 

[Gatos]

audacity and audition (and probably every good audio editing tool) have a center channel extraction filter. And for that a stereo track is needed (yes, that is 2.0). The center channel is what comes equally from both channels, so it extracts what is mono. This is very often true for the voice track and for sfx, so with this method you can filter out vocals (or at least a lot of them) and sound effects.
Whatever you want to remove from an audio track, this is no exact science. If voice and sound effects and music are mixed into each other, you simply cannot separate them 100%.
But you can get close with various filters.
Here is the center channel extraction method explained to GET the voice track:
1. load the stereo audio track into your audio editing tool.
2. use the filter "center channel extraction" (adobe audition comes with a few more filter goodies like setting the frequency for voices, which can be very helpful)
3. mark the result
4. use the filter inversion to do a 180° flip of the wave structure.
5. copy the sound file.
6. by using the "reverse" function go back to your original unharmed sound file.
7. mix the copied track into the original sound file (at 100% volume).
8. the result should already show clearly that parts have been removed from the original fie. Now listen to it and if this got you any closer to clear voices without music, then you can continue working from there.

Thanks boon. I'm working with stereo and I want to extract voices and remove background music stuff.

 

[nyl]

generally speaking, is it easier to remove the music from a scene when done using the original 5.1, or is it easier (in general) to do so after using headache to output the 2 channel pcm? Noob here to audio editing using 5.1.

 

[boon23]

generally speaking: there is NO easy way. Try out whatever works or even gets you a bit closer to voice extraction. The described method however requires a stereo track.

 

[nyl]

ok, have the stereo from the 5.1 downmixed. No worries, i will mux it up. You guys are all Hardcore, thanks.