Audacity Forum

I've been wondering alot about why the Audacity-team wont let us use this effect the other way around as in keeping the center-channel but removing the lefts and rights. The point for me with this tool is just to be a little able to seperate the sounds on recordings. I use this effect as a tool, to sample sounds from proffesional artists that i like, but it really does annoy me that i can't keep just the center-channel.

Secondly i've been wondering why they call this thing a vocal remover anyways, when really it's a center-channel-remover!??!!? It's a very misleading name for the effect. I applied it to a Michael Jackson track, and i was left with just the vocals, cause apparently he likes to pan his voice around, while keeping all the drums and rythms in the center. This is kind of an ironic result when applying an effect callled a vocal REMOVER!

morob wrote:I've been wondering alot about why the Audacity-team wont let us use this effect the other way around as in keeping the center-channel but removing the lefts and rights.

It's not the Audacity team that stop you from doing it, it's the laws of physics.

Take a simplified version of how it works:
Lets say that
the sounds that are panned hard left are "a"
the sounds that are panned hard right are "b"
and the sounds that are panned centre are "c"

Then the left audio channel ("Left") contains "a" and half of "c"
and the "Right" contains "b" and half of "c"

If we invert the Right channel we have "minus b" and "minus half of c" (-b -c/2)
and if we then mix the two channels we are adding this to the left channel (a + c/2) we get "a +c/2 -b -c/2"
"c/2 - c/2" gives us nothing (they cancel out) so we are left with "a - b" (the left channel and the inverted right channel).
We have centre pan removal.

We can do it the other way round, invert the Left channel and mix it with the Right, then we get "a + c/2 - b - c/2" which gives us the inverted left channel and the non-inverted Right channel (b - a)

It'll drive you mad trying to work it out, but there is no way to just get "c".

morob wrote:i've been wondering why they call this thing a vocal remover anyways, when really it's a center-channel-remover

It's a "tradition" thing. This effect is traditionally called "vocal remover". Old stereo recordings usually had the main vocal panned dead centre, though it is less common in more recent times as the fashion for "highly produced" recordings have become fashionable. Add to this the fact that MP3s will frequently mess up the stereo separation, and we are left with a situation where "vocal removal" often does not work.

In the latest (1.3.8 alpha) version of Audacity, the effect is called "Vocal Remover (for center-panned vocals)...", which I guess is a better compromise (though I wish they spelt "centre" correctly - but that's the official language convention for Audacity).

Thanks a lot for the reply. I'm not gonna pretend that it makes sense to me, but i trust you when you say it can't be done. It's a shame though, cause there's often a lot of very cool drum-sounds right in the middle, that i'd love to steal!

Oh wait i think i'm gettin it a little. Is it that the inverted wave would cancel out the non inverted wave if you put the two waves on top of each other. And then this somehow only works for sounds that have the exact same wave on both the right and the left channel??

The other tool is called Vocal Isolation and there re certainly tools that can do it. Some that do both.

Voice Removal or Isolation
--Center Pan Remover (Voice Remover--Search the page for "Center Pan Remover")
http://audacityteam.org/download/nyquistplugins
--Voice Trap
http://www.cloneensemble.com/vt_main.htm
--Extra Boy
http://www.kvraudio.com/get/1651.html
--Kn0ck0ut
http://www.freewebs.com/st3pan0va/

A good place to start is the assumption that you can't remove vocals from a show. Can't be done. Sometimes, under rare circumstances, you can remove some vocals badly from a crisp, super good quality, stereo show, assuming the singer is in the middle and nobody added dimentional effects, stereo reverberation, or echo to the performance.

Any musical instruments in the middle will be affected, too.

Given those restrictions, there are about five songs ever recorded that make good candidates for vocal removal or isolation.

What most people want is to remove vocals from a downloaded, ratty, mono MP3, or worse, MP4 or M4A.

Sorry. Can't do that.

Koz

Well what i really wish to do is to be able to split all the sounds from eachother, and what i'm using the tool for really has nothing to do with vocals. It rally does work on a lot of tracks, like for instance a lot of trance and psytrance music will have the kick and the main bassline right in the middle, and snares and claps slightly panned, and hi-hiats slight to to hard panned. Theres a lot of recurring themes in this music where all the pads and leads are muted, and the kick and main bass will do a solo alongside some other percussion. On these parts it is possible to apply this effect and extract some great sounding profession percussions. In the future perhaps they'll invent advanced technologies that can pick out any single sound from a mix no matter where it's panned and even seperate the effects from the RAW sound. I mean it must be possible to achieve this, since i'm achieving it inside my head when i hear music. The technology just doesn't allow for this yet...

<<<On these parts it is possible to apply this effect and extract some great sounding profession percussions.>>>

Which one of the tools did you get to work and did you apply it in any special way?

Forum = Dialog.

Koz

steve wrote: It's not the Audacity team that stop you from doing it, it's the laws of physics.

Sorry to resurrect an old thread, but I must respectfully disagree. As far as chasing down a vocal track that moves around, I can't help you, but I needed to isolate a track that was very low in the mix (to score it) and was mostly successful. If the track you want has already been mixed center panned, you got it easy.

EDIT: I figured out a simpler way to do this, so I've revised these steps

1. Import your stereo song.
2. Merge them (Tracks/Stereo Track to Mono). Let's call the resulting track "a".
3. Import it again (File/Import/Audio... and select the original stereo audio file again). You should now have a total of 3 tracks (well, one mono and one stereo).
4. Split the stereo tracks.
5. Select either the left or right track and use Effect/Invert.
6. Make the L & R mono tracks stereo again.
7. Select the stereo track and merge them now (Tracks/Stereo Track to Mono). We'll call this track "b".
8. You most probably need to invert either a or b now (not 100% sure what the result will be all samples). Zoom in on a & b to make sure they have opposing polarity.
9. Now make mono tracks a and b into a single stereo track.
10. Merge the remaining tracks.
11. Normalize

Alternately, you can replace steps 4-7 with the vocal remover plug-in, but you'll also have to split the track and delete one of them (since they will be the same). The vocal remover plug-in appears to leave you with the inversion of the original (minus center-panned audio) -- this may vary depending upon the version of the vocal remover, so you may need to zoom way way in and make sure that the two channels are of opposite polarity before you merge them (i.e., the lines are on opposite sides of the zero bar).

I was able to use this method (after 1st attenuating the left track by about 40dB) to mostly isolate the track I wanted. If I get my panning perfect, I think I can isolate it more, but it's good enough for what I need to do (I've essentially amplified the instrument I wanted by around 20dB or so)

I'm a good C/C++ programmer so if I get unlazy enough, I'll write a plugin that does this for you. Just don't put your life on hold to waiting for it

Thanks to all of you devs for making this available!!

Daniel

EDIT: In case I didn't make this mathematically clear enough, what I'm saying is essentially this:
where:

• r is right channel
• l is left channel
• c is center panned audio
• o is all audio other than center panned

Center panned vocal remover works like this:
o = r - l

Therefore:
c = (r + l) / 2 - (r - l)

It does remain that many of our posters want to do this with a mono show or a show that's been highly compressed or a show that's been highly produced. Those shows do not lend themselves to post production isolation/cancellation no matter how much math you throw at them.

Now I'll sit back here and watch this proceed. There was a poster here a while ago that was able to do a reasonable job with isolation and rejection using his own proprietary processing and software.

He, too, was working with a perfect quality stereo show.

Koz

daniel.santos wrote:Sorry to resurrect an old thread, but I must respectfully disagree

No problem, people often do disagree until they've exhausted all options.

daniel.santos wrote:1 Import your stereo song.
2 Merge them (Tracks/Stereo Track to Mono). Let's call the resulting track "a".
3 Import it again (File/Import/Audio... and select the original stereo audio file again). You should now have a total of 3 tracks (well, one mono and one stereo).
4 Split the stereo tracks.
5 Select either the left or right track and use Effect/Invert.
6 Make the L & R mono tracks stereo again.
7 Select the stereo track and merge them now (Tracks/Stereo Track to Mono). We'll call this track "b".
8 You most probably need to invert either a or b now (not 100% sure what the result will be all samples). Zoom in on a & b to make sure they have opposing polarity.
9 Now make mono tracks a and b into a single stereo track.
10 Merge the remaining tracks.
11 Normalize

That will either remove the left panned audio or the right panned audio depending on what you do in steps 5 and 8.
Here's a stereo sample to test with: It has a sine tone panned left, white noise panned right and a click track panned centre.

Depending on the audio that you start with, eliminating either the left or right panned audio may make the centre audio easier to hear.

daniel.santos wrote:I'm a good C/C++ programmer so if I get unlazy enough, I'll write a plugin that does this for you.

This can already be done using the "Channel Mixer" plug-in, but a C/C++ version would be nice as it should be possible to make it faster, and enable "Preview".