I am trying to clean up noisy LP selections. The problem is a frying, crackling noise throughout. Noise removal does not work satisfactorily due to the randomness of the noise, which typically consists of more or less triangular spikes roughly 3-6 high and maybe 10-20 samples wide superimposed on the original waveform. I cannot find settings for click removal which work either.
What does seem to work is the repair function - for a single instance of a crackle, if I isolate it visually and apply it. But there are literally tens to hundreds of thousands of these in the album, so this is not practical.
Is there any way of automatically iterating the repair effect over a large sound selection by setting the “repair window” within the effect’s limit and a “window start increment” to move the repair effect iterations through the selection?
I am using Audacity 2.0 on 64 bit Windows 7 Professional, installed from the .exe installer.
Don’t know if i really understand what you are saying but, once you have selected the noise you want to remove you then click effects> noise remopval, and the do a profile and then select all of the recording, and follow the instruction there, do not click repair. hope this is waht you were talking about, if not please excuse the interruption
No matter whether I select the “quiet” passage before the music begins, or select a single tick in the crackle for the noise removal profile, noise removal fails.
Here is a sample showing a few of the crackles.
What I am suggesting (and this is for the developers of this product for a future enhancement) is this: Given a sample (could be the entire piece) with beginning b and end e, specify and repair window of width w and and increment of i. Do interated repairs from b to e-i on the waveform from b+ni to b+ni+w where n is the iteration index.
Here we would have b set to 0.984, w=.002 & i=.001. Run the repair through a few of these increments and you get
The main problem there is that the repair effect would seriously damage audio that should not be repaired.
For example, the upper track here is normal (undamaged) audio, and the track below is where it has been (manually) iteratively repaired. You will notice that the transient peaks in the first half have been destroyed by the “repair” process.