What is the best way to do this other than manually?
I feel like an intelligent algorithm and AI could take care of this pretty easily. If the user tells the program the BPM, and the event is properly in time on the grid/adjusted, the software could know to look for those metronome frequencies on the beat and avoid removing areas of differentiation found throughout the whole event.
I canât tell whether youâre offering advice or looking for help.
Have you tried using the declicker? Iâve never heard of a declicker algorithm that didnât cause at least some artifacting, however, breezy talk about âintelligent AIâ notwithstanding. Precision, unfortunately, usually requires doing things by hand.
The de-bleed is good for this. But you would need an audio file with the click only and at the same length. Of course, you could grab a sample of the click from your audio and create a new audio file and paste that click into it and make copies that line up with your music track then run the de-bleed.
Apart from the risk of some small artifacts persisting, âSelecting similar patternsâ as @MrSoundman says above, should get pretty close to achieve what the OP wants⌠?
My suggestion should work, but obviously takes some human work to set it up.
A few minutes lining up copies of clicks against the infected file shouldnât tax the people here