It sounds like a heartbeat. Here’s the audio sample. https://drive.google.com/open?id=18h_HtwMWlBkgee8tSLA_4ciEWUwb_NJR
It sounds like a heartbeat.
Not only that but the rate is almost exactly half a human heartbeat. 36 > 72
What’s the goal? Theatrically pure recording for sale or sharing? I don’t think you’re going to get there. There’s just enough splatter across the audible range to make normal filter tools not work right without damaging the voice.
You didn’t shoot it, right? So you don’t know what they were doing during the performance. Sometimes that can point to a solution.
Desperation Method might be to copy a bunch of the thumps-only onto their own timeline and use that in Effect > Noise Reduction > Profile. Then Noise Reduce the whole lecture. It’s really labor intensive and it may not work.
https://manual.audacityteam.org/man/noise_reduction.html
It could be a plain, simple method of Copy Protection. Where did it come from?
Koz
Selecting and suppressing each thump doesn’t work, either. That leaves you with either holes or an annoying tick instead.
Koz
Not sure of the source of the tape, it belongs to a friend. I’ll just leave it alone. Someone was able to remove heartbeat sounds with delicker, thought this could be done too but I guess there’s no easy way to do it.
https://forum.audacityteam.org/t/how-to-remove-the-beating-heart/49523/1
You can substantially reduce the spikes by filtering out everything below 200 Hz with the “Filter Curve EQ”
(Note the vertical sliders on the right hand side)
filtering out everything below 200 Hz
I tried it at 300Hz but it sound too much like a bad telephone.
Koz
@steve Thanks that works well and better than declicker which took a long time. Did you determine the noise is 200MHz using spectrogram?
I zoomed in on a couple of clicks and saw that the spikes were quite broad (low frequency), so I tried a quick test with the “Low Roll-off for Speech” preset. That preset is designed to remove as much low frequency as possible without damaging speech. It made an improvement, but only about 60%, so I then looked at the track spectrogram view (https://manual.audacityteam.org/man/spectrogram_view.html) and zoomed in vertically (https://manual.audacityteam.org/man/vertical_zooming.html) on the low frequencies.
The low frequencies in speech typically go down to around 150 Hz.
Landline telephones generally go down to around 300 Hz, which is enough for ineligibility but lacks the deeper timbre, especially for deep male voices.
The spikes in the spectrogram appeared to be mostly under 200 Hz.
So a cut-off, as steep as possible, at around 200 Hz looked like a good compromise.
I tried some of that, but basically guessed at it using the telephone as a model.
Koz