10 bands over a region of 1kHz was just for that bad 4kHz whistle.
10 bands over 10kHz range from 3 to 13kHz is more normal for de-essing.
Re: Spectral Editing:
Whistles aren’t always at a constant frequency , so sometimes a notch-filter won’t remove it all …
whereas your de-esser will remove a whistle of varying frequency.
Photoshop-[or GIMP]-style editing of the spectrogram would be a dream come true : dodging / burning / erasing / contrast-control etc using brush strokes.
We can dream about the dodge and burn, but selecting simple rectangles and doing something with them is a good advance to begin with.
Ten from 3 kHz to 13 kHz makes wider bands (2.5 steps) than with my defaults. I think those setting treat that whistle well enough, but I may not have the keenest ears.
I just treat from 2.5 to 8 kHz. Certainly not above 11.5 kHz which is a cutoff for even the highest quality of audiobooks downloadable from Audible, as I have learned from some experiment.
But my spectral editing can do other than a notch filter. It can do a parametric equalizer. The de-esser relies on the same eq-band function of Nyquist.
Re: DeClicker … it’s also a DeCrackler if you push it to the max …
Now the bad news …
processing to DeCrackle takes about 10x the playback time on my 5 year-old dual core computer,
(your mileage may vary).
It seems then that this too should be verified by Steve then documented by Paul and put out to Audacity Wiki. Or does it need more development?
Gale
Which extreme settings were you using now?
My intention was to write something that removes naturally occurring but distracting mouth noises from narration. It was not my intention to fix the noises of damaged old media. Fixing spikes in a signal may actually be a different problem from fixing small natural sounds of short duration. But, if you think it works, you’re welcome!
As it happens I was just lately looking around for good books about Audacity for instructing other people, I don’t know if you have recommendations, and I found one just about fixing up old 78 rpm’s, which is free if you have Kindle Unlimited. I think there are some interesting ideas here.
DeCrackle looks like the DeClick problem but on a smaller time-scale, so a DeClicker will also DeCrackle, but inevitably involves a lot more computational effort because the occurrence is more frequent.
The lower limits of the “repair interval” and “precision” maybe need to be lowered for DeCrackling because of the shorter time scale compared with DeClicking.
The DeCrackle sounds smoother than paid-for software like Brian Davies “click repair”, see attachment ,
( but to be fair the processing-time is easily 20x longer than Brian Davies ).
Did you reduce the minimum separation permitted between clicks? It is undesirable to put it much lower than the default, I find, for editing my own male voice. It would muffle too much. As it is, it muffles the vocal fry a bit, that ends of sentences sometimes trail into.
Here’s a crunchy one from 1923 …
The settings are a compromise : trading off DeCrackle against processing time.
The settings shown give a processing time of about 4x playback time on my computer.
You pushed the repair interval down to 0. I thought I needed nonzero repair intervals for crossfading of each fix so that I do not introduce clicks by making edge artifiacts at each interval.
If you did not reduce repair interval, you might not need so many passes to compensate for the clicks created by earlier passes.
So you might find a more efficient set of settings for this purpose.
Musical applications of Paul-L’s DeClicker & DeEsser…
Hello! Sorry for the thread necro but it seemed like the best place to ask this and may help other would-be declickers.
I have been using your plugin to remove some mouth clicks that I can’t seem to get rid of naturally. At first I couldn’t get it to work at all but then forum user Trebor (super awesome guy!) helped my greatly by suggesting some guideline settings that removed the clicks completely. However, the process took prohibitively long. I record youtube videos that are 20-30 minutes long.
So over the past month or so I have been trying to tweak it to lower the computation needed but still remove the clicks. Thing is, I’m not an audio guy. I don’t really understand the terms, and so my process has just been to move numbers around and see what happens - not very intelligent.
The settings Trebor came up with are:
However, he also attached an animated spectrogram which appears to me as if the clicks I am experiencing are from 800/1000 and above.
So I would assume I could still remove the clicks and save processor power by changing the frequency from 800-16000.
Also, he chose threshold 3 over the default of 5. I understand the higher the threshold the faster the scan will go, so in layman’s terms what’s the difference between 3 and 5, or 4, 2, etc? Also in layman’s terms, what is the practical effect of 1 pass vs. 2? If the other settings are the same, how could a 2nd (or 3rd) pass catch something the first would miss?
Finally, and I may be asking too much (desperation lends itself to this I’m afraid) are any of the other settings editable, and how so, to save time but only minimally effect the positive result?
The link to my original thread is: https://forum.audacityteam.org/t/clicks-beeps-in-recorded-audio-w-video-examples/35818/1
Thank you so much for this helpful plugin and your time!
-Marcus
Hello Marcus, a few pointers.
I never adjust the threshold now. I leave it at 6 dB. I find the way to get more clicks out is not to lower the threshold but to increase the number of frequency bands. I might use the maximum number of bands on very short selections for spot treatment of the more resistant clicks.
The computation does get more expensive as bands get more numerous and narrower. I don’t bother treating above 10 kHz where human hearing is not very sensitive. But it is also true that lower frequency bands are more expensive to compute than higher so that 10 kHz+ range might not cost very much.
Peak hearing sensitivity is between about 1 and 5 kHz and if you want the best results for limited computation time, try concentrating on that range. Perhaps eight bands.
I also treat files of voice that may go to thirty minutes, and yes it is slow and I am patient for the results. I do like to include bands that go down to low frequencies. There is a sort of low-frequency click sometimes under high frequency sound, the opposite of the usual pattern. Rather than the clicks inside vowels, I mean the rattles that may happen inside s sounds. The declicker removes most of these and I like that but it does entail the expense of treating the low frequencies. Maybe you are willing to dispense with that.
By the way, if you are accustomed to looking at spectrograms, the log f spectrogram scale may be more helpful for understanding the declicker settings (also the deesser settings). Bands have equal height in the logarithmic scale, not the linear one. The width of bands would be better described by octaves, not Hertz. In the default settings, 150 Hz to 9600 Hz is six doublings, that is six octaves, and 12 bands mean each band is half an octave.
This is not exotic if you are also familiar with Audacity’s graphic equalizer, which also divides the logarithmic scale of frequency and has three sliders per octave.
Are these two tools being developed further?
Will there be a linked stereo de-esser?
Gale
Paul,
Thanks for the reply. I’m actually not familiar with spectrograms at all, I just was able to understand the one Trebor posted, as you can see the clicks being removed right in front of your eyes.
If I get what you’re saying though, It’s that all else equal, 3db or 6db threshold will remove the same number of clicks. Rather, the number of bands is the most relevant to removing more clicks. As above I used 20 bands and it removed everything. You suggest 8 bands. Any significance to that number? Also, for clarification, anything between 10,000hz and 16,000hz cannot be heard by a human? That’s a pretty big processing drop so I’m very happy to know that. Also, would you please clarify the relevance of passes and the benefit of 4 or 2 vs. 1. I tried a file last night, 50% one pass, 50% two pass, and both halves came out click free. That being said, maybe I just click more in the first half (when I did the two pass). Again, I apologize for the total lack of audio terminology knowledge, but in my head I’m justifying asking by figuring many folks who use Audacity - at least in the Youtube crowd - use it paint by numbers and would be helped immensely, like me, by a more end-user definition.
Thanks again!
-Marcus
Many questions there.
What I use Audacity for is production of Audiobooks, and audible.com does some of their own audio processing and data compression. I have learned that even in the highest quality retail versions, data compression apparently throws away everything above about 11.5 kHz. So I don’t worry about sound above that range.
Multiple passes might remove a few more difficult clicks for reasons I won’t detail. It involves cases when clicks come very close together. But I expect there is diminishing returns and I never use more than two.
Lowering the threshold, in my experience, is not a way to remove more clicks, and it might instead make too much unnecessary change to the waveform, changing parts of the sound that are not clicky and also wasting more time in computation. If you use the Isolate choice in the drop down, then instead of repairing the sound, you can see and hear what would get subtracted to make the correction. You don’t want so much “murmur” coming through that you can almost understand the words. That seemed to be the case in the three way example Trebor posted in the other thread.
I’ll say some more. The controls let you define a top and bottom frequency and a number of bands. What is the width of a band? Use this formula:
ln (top / bottom) / ln(2)
And that is the number of “octaves.” With my defaults of 150 and 9600 you get exactly 6.
So with 12 bands, each is 1/2 octave, or 6 neighboring piano keys (including the black ones). With 18, each is 1/3 octave or 4 keys. (like the sliders in Audacity’s graphic equalizer). With 36, 1/6 octave or 2 keys.
I could have made the controls for bottom frequency, individual band width, and number of bands. Perhaps that might make more sense to the user. Or not.
If you want to treat only a part of the frequency range, say, ignoring low frequencies, but treat the remaining middle and high frequencies with the same thoroughness, you want each band to be just as many piano keys.
Say you treat 1000 to 5700 Hz, where peak hearing is – that’s close to 2.5 octaves. So 5 bands would be equivalent to the default 12 bands with the default top and bottom. But prefer 18 bands now, so with this narrower range, that would correspond to 7.5 – we can’t have a fraction, so say 7 or 8.
Thanks Paul! I won’t ask any more questions! I will check for murmer as I agree with you I don’t want to muddy the sound. I tried one today with 6 threshold, 800-15000, and 2 passes, 12 bands. I found no clicks, but I did find a couple deep sounds where clicks might have been, possibly that low band psuedo click you mentioned. It didn’t bother me though, I just cut it out as it was during a silence.
The whole thing took 17 and a half minutes for a 21 minute video so I would call that a success!
This plug-in is a lifesaver for my sanity, so thank you!