I’m fairly new to all audio editing, and I’ve searched these forums for guides on removing unwanted noises, but many seem to be application specific as opposed to general frequency analysis.
By way of example, I’m working with the attached track which is a 20 second segment of a pixabay copyright free song “Uplifting Piano ls”.
Most noticeably at t = 5.2, 10, 16 s you can hear a popping/click noise. If I zoom in to these segments, I can see jagged-looking waves at a much higher frequency occuring. However, they don’t all seem to be at the same frequency, so applying a notch filter doesn’t eliminate them all, and turns into a game of whack-a-mole. I can low pass the entire thing, but that ruins the overall tone as some of these “pop” noises occur within the range of otherwise desirable overtones.
Surely, there must be better ways of identifying undesirable noises; defined as if > 99.99% of a song plays at certain set smooth frequencies and the rest are outliers. When I run a frequency analysis, it shows the expected peaks of the core treble melody of C5-E5-G5, and then a bunch of high frequency stuff, too numerous to identify the problem.
If someone more experienced could lend any tips to zero in on unwanted frequencies and eliminate them, I’d greatly appreciate the help.
If they are short-duration (like I think) you can try the Click Removal effect (automatic) or the Repair effect (manual, where you have to select each defect individually) or you can zoom-in and try using the Draw Tool to re-draw the waveform.
There’s little clicks and pops all though that performance. Some are worse than others.
There is one desperation method. The tune seems to have repeating phrases, Pick an undamaged note and lay it over the damaged note in the mating phrase. Repeat as needed.
I don’t think there is a “push this button” type solution.
This is where the person wearing the Producers’s Hat starts planning on what happens when the tune doesn’t come back.
Indeed, I did notice that as well - the quality is poor throughout.
What’s odd is this is just on the left stereo track (the segment I uploaded is only the L), and the right is fairly clean - I would have thought whoever recorded it would be capturing it at the same time.
I do like the undamaged segment method, though I often find this leads to way more time than expected patching up the transitions, and even though the casual listener might not notice it’s like an itch to be scratched.
Appreciate your contribution though, as for hats I think it’s time to don the Businessman’s Hat, and simply find another track without recording errors that suits the mood in my application…
@DVDdoug I did try both of those tools, the click removal was ineffective over a range of parameters, and the repair WAS effective, however simply too cumbersome; many of those “click” sections you hear consist of maybe ~400 samples, and the error sections would be interspliced with ‘good’ sections, causing tearing of the waveform unless the repair tool was used extremely judiciously. Point being it was more laborious than I was willing to spend on it, but I appreciate the suggestions.
Those tools; click removal, repair, then low-pass filters and notch filters have been my go-to tools, so I may be missing something, eg ive never used the noise gate. The Spectrogram/Multiview is certainly a beast I need to learn.
The pops sound like digital artifacts created by AI, e.g.
(their envelope is like a reversed, i.e. backwards, sound).
Re: “copyright free” if a person created it using AI they may not have the right to waive copyright.
The copyright of AI generated music is a grey area.
The AI software maker can claim copyright,
and the artists whose work was used to train the AI can claim copyright, if they did not give permission.
