I have several musical tracks, one clean and complete and the others are fragments of the clean track mixed with sound effects like (wind, water…)
Is there a way that audacity can automatically align the fragments with the clean track?
Is there something like booleans operations (AND , OR , XOR , NOT…) for audio because I want to keep the audio that is almost identical to the clean track from the fragmented tracks. So it would be the operator AND here in this case and Subtract would be NOT.
I know this may seems useless and silly if I have the complete clean tracks but I have my reasons.
I encounter the same problem when I want to unmix something and I used to think It is almost impossible to do in a free program with free plugins.
I guess that in paid pro sound solutions there is a way to even isolate a particular voice or instrument according to a clean sample.
When I want to clean a musical track from unwanted noises I used to use vocal remover plugins then equalization. It works when the noises or either
the music are center panned. The problem with vocal remover is that you can’t select a particular point of panning to be isolated or removed.(I guess It’s not technically possible tough.) And the problem with equalization is that you don’t have a curve of intensity (0-100%) of the effect according to time(selection) in order to make the effect seamless in the whole track.
I was wondering the possibilities of a free program like audacity toward the paid counterparts. I hope my post is understandable.
There is no “almost identical.” to get any of these tricks to work, you have to have top quality, uncompressed, identical tracks. Most people who post on the forum are trying to do tricks with MP3 or other compressed sound files that even just listening to them aren’t identical.
A person listening to the tracks “knows” what a singer and piano are. The software has no such intelligence. It’s seeing just a bunch of sample points. This also kills people with Noise Reduction. It’s obvious to you because you “know” what noise is. The software has to be taught and guided.
Center Pan (Vocal) Remover is the only process of dead simple arithmetic. All the others (including center pan isolation) are highly complex tasks that try to “know” what individual performers sound like.
I guess that in paid pro sound solutions there is a way to even isolate a particular voice or instrument according to a clean sample.
Well a program that can analyse and isolate instruments automatically no and It is highly doubtful there could be one.
However I now this one but you to do a lot manually which is not a problem and I guess you know this program:
I would definitely buy this software if I was a pro and or have 340 euros to spend on a software for regular use.
It seems there is another program by Izotope but I haven’t checked It yet.
I think I have a noise removal tool plugin with audacity with sample but if I remember it doesn’t work or is disapointing.
Do you know such a plugin that is satisfying enough ?
I guess from what you said that there is no way to automatically align tracks according to non exactly matching content or (beat).
But what about the idea of a curve of intensity for effects plugins that I described earlier ?
Or even simpler a way to fade in and fade out the effect. Does it exist already ?
Maybe It’s a feature suggestion.
if I was a pro and or have 340 euros to spend on a software for regular use.
They know what they have.
I think I have a noise removal tool plugin with audacity with sample but if I remember it doesn’t work or is disapointing.
Do you know such a plugin that is satisfying enough ?
There is significant work on new Noise Removal software for Audacity. As part of that upgrade, the name will be changed to “Noise Reduction.” Too many people expect it to actually remove noise. To zero.
But what about the idea of a curve of intensity for effects plugins
Good idea, but I know what the come-back line is going to be. You can work around that with clean and effected tracks and custom fading with the envelope tool.
A lot of the effects work is going toward fixing obviously broken tools, filters and effects. The rest is pushing out the latest Audacity. Mac Yosemite in particular has been a real joy to troubleshoot.
There is also “Melodyne” (if your pockets are deep enough).
Within certain technical limitations, it is possible, though it is difficult to provide an easy to use interface for the task.
If the copy of the sound that you want to remove is identical (on a sample level basis) to the sound that you want to remove, then it can be “subtracted” by inverting, scaling to the appropriate level and panning to match the original.
Example 1:
Track A (original) has unwanted sound in left channel only.
Track B has the (identical) isolated unwanted sound (mono).
What you would do is to make Track B into a stereo track (2 channels) with the sound, inverted, in the left channel, and silence in the right channel. Then mix (“add”) Track A with the new Track B.
Example 2:
Track A (original) has unwanted sound centre panned.
Track B has the (identical) isolated unwanted sound (mono).
Invert Track B and mix with (“add to”) Track A. Audacity will automatically convert Track B to stereo because it is being mixed with a stereo track. Thus Track B, when inverted, is the negative of the unwanted sound.
Similar operations can be performed when you have an identical, isolated copy of the sound that you want to remove, wherever it is panned in the original stereo track. All you need to do in this case is to find the necessary pan position and scaling factor (though that is easier said than done).
When working with sounds that are not identical at the sample level, then the task becomes very much more difficult. However there is an FFT based plugin called “Stereo Tool Kit”: Karaoke, Rotation, Panning & more
Installation instructions are the same as for other Nyquist plugins (see previous link).
Note that this is an experimental plugin from Robert J H. Notes about how to use it are in that topic - if you need further information about that plugin, please post to that topic.
The next version of Audacity includes some simple “spectral edit” tools that allow boosting and cutting parts of the sound based on frequency, using the track Spectrogram views.
Steve has mentioned my 2D Stereo Toolkit.
It has an additional rotation parameter when applying center removal or isolation.
For example, if the voice is halfways between center and right speaker, you’d enter -22.5 (degrees) to bring it into the middle before extracting.
Regarding intensity:
I assume that you mean some kind of envelope that acts on the effect itself, or am I mistaken?
Supposedly you wanted to remove the center only in specific passages, then its straightforward with the 2DST:
Duplicate the original
apply "Center Isolation (inverted).
Played together, the center will be gone.
You could now use the envelope tool to change the amount of the effect (or gain sliders for the whole track):
on the second track, the center comes back when you drag the envelope down.
on the first track, the side channels are attenuated, you’ll prominently hear e.g. vocals, bass and drums (like soloing the second track).
There’s a sample file in the topic that shows just this.
numerous effects can be treated like this–duplicate the track, apply the effect and invert it if necessary (reverb would work without that).
Some effects produce a delay though–you’ll get a comb effect when playing the tracks together.
Hello. I am sorry I was away and was unable to reply and sorry to bump this.
Good plugin, so it seems the panning can be rotated and then the new center can be removed?
Thank for the tips.
Effect parameter = (stereo field) rotation range [-180;180]
I guess I have to tweak this in order to rotate the panning but does it have to be exactly accurate to be effective ? I can approximatly determine by ear but checking the whole 360 degrees range seems tedious.
Is there an analysis plugin to find how things are panned not just by ear ?
There’s a fixed rotation option in the plug-in (probably called carousel or something)
The idea is to rotate the audio e.g. 1 degree per second and then to remove the vocals in a second plug-in call (delete the parameter “1.0” this time).
Listen through the track and try to make out where the vocals are attenuated the most. The time index in seconds gives you the rotation parameter to enter (after undoing the last two plug-in applications, in order to have the original back).
However, the rotation is not a “true” one. The more rotation, the more inverted phase will be present at the extreme pan points. That’s the nature and relationship of stereo sound.
You can also pre-pan the audio to extract another center.
Use the pan slider to shift the target audio into the center and then render the audio before the 2DSTK.
There’s currently no analyse tool available for this kind of stuff, I’m afraid.