Did you save originals and compare before and after? Were the “deep sounds”, the bumps, there before or is my declicker introducing them?
I did not. There was only a couple of them and they were easy to remove as they came in dead zones. Much better than the literally dozens of clicks. Your tool is a godsend. Funny thing, I used it a couple nights ago and it was over in like 5 min, for a 25 min file. I didn’t change the settings, which is normally about 30 min for such a file, and it appeared to work just fine - no errors, no clicks. Maybe my computer just had an elite moment ![]()
What settings have you settled on then? Knowing this might help other users.
I run default, with the following changes:
Range changed to 800-15,000.
Bands increased to 15.
I think what got rid of my specific clicks from default was the expanded range. I have seen good results with 12 bands, and equal if not better results with 20.
Edit: I kinda wish that it would remember my settings though like other plugins such as normalize, compressor, and equalizer.
-Marcus
If clicks above 9600 are audible to you, your ears are keen.
Raising the bottom to 800 might account for much of the speed improvement. Lower frequencies take longer to compute.
If you want to change the default settings, that is easy to do with a text editor. There are lines near the top of the file that begin with ;control specifying minimum, maximum, and default values for each numerical slider.
Nyquist plugins can’t do that yet.
Gale
Paul,
Random question you might be able to answer. If I can use the setting above to de-click a 24 minute video in 30 minutes, why would it take about 3 hours (or thereabouts, I’m still waiting) to do a 26 minute one? What, aside from the settings and the length, could increase the processing time? Could more clicks take longer, or does that even matter?
Thanks!
Some questions for Paul L with a view to making these plugins available on the Audacity wiki.
If/when these plugins are “ready”, would you like them to be published on the Audacity wiki?
Are the versions in this post https://forum.audacityteam.org/t/updated-de-clicker-and-new-de-esser-for-speech/34283/1 the current versions?
Would you describe these plugins as “release ready” or “experimental”?
If “release ready”, can you point me to any documentation that there may be?
What license is the software under?
+1 to that unless there are bugs.
Do I assume the de-clicker being slow at lower frequencies cannot be improved?
Gale
Are you certain there was no difference in settings? I am really not sure what causes an effect that great.
I have seen that the time it takes to process audio is not proportional to the length of the audio (when I have had no reason to think the frequency of clicks differs). But I don’t think the curve has ever been that steep for me.
But who knows. This might be very dependent on the amount of memory in your system. I have an unconfirmed suspicion that the garbage collection of the Lisp interpreter for the programming language that implements this effect is involved.
Thanks and congrats for soon being on the Audacity wiki!
Ohmigosh, you are a genius. I am so grateful for these new tools. They are wonderful. Thank you for being brilliant and for sharing it!
This is brilliant. I’d like to suggest a new name for your De-Clicker to differentiate it from the Click Removal effect used for vinyl recordings. How about “De-Smacker”? As in smacking lips… Okay, maybe not. These tools, however, are totally brilliant. Thank you for sharing your genius!
We should persuade the author to document and release them on http://wiki.audacityteam.org/wiki/Nyquist_Effect_Plugins - ideally with a linked stereo option for De-Esser.
Gale
I want to research better ways to do this that don’t take so much compute time. But Gale knows I have too many other ideas I am chasing at the same time.
What’s the “this” there, Paul? Do you mean you think both effects are too slow to publish on the Wiki yet?
Although certain standards are needed to publish on the Wiki, ultimate perfection is not required. At least enough description of what the controls do is required.
Gale
For many activities like this the utilty of the function is often far more important than the compute time involved.
The important thing to do is to advise the user up-front how long the function might take on their currently selected audio, so they can decide whether or not to press ahead with it - and for longer ones they can make a cuppa or do the housework ![]()
Peter.
I agree completely. Better to have an effect that is slow, than to not have it at all!
The user can decide. And the more it gets used and gets attention, I think the more motivation there will be to speed it up.
-NTL2009
Sorry to necro this thread a second time!
I found something that helped me immeasurably with processing times! My audio recorder that I use the most for gaming capture forces you to record in stereo. Since it’s just me talking into a mic, stereo is unnecessary and so I run the tool “covert stereo track to mono” first thing in Audacity. This cuts the de-clicker time MORE than in half because not only does it take half as long, but there is a period where things hang for a few minutes between tracks (L/R, that is) and that goes away in mono. Now I can do a 30 min episode on about 10 de-clicker minutes! Wooo!
But Paul, now that I’m more experienced I’m trying to get even better. Based on your earlier notes it would appear that as you raise the frequency bands that you get better results (and it’s the only tweak you recommend - you recommend to NOT to decrease the sensitivity threshold). I have not found this to be the case. It seems in my case 16 or 17 removes more clicks than 18, which is an odd result. Since you mentioned there are bands within the octave, so maybe you need certain multiples?
In essence, I’m getting a lot of the clicks out, but not all, especially when they coincide with a “high note” a bit of sound I am emphasizing, i.e. my voice gets a bit louder and sharper than the rest of the word. Just today I noticed it twice in a row on the “a” sound when I said the word “brain” - the closest I have got to a “repeatable” click. My guess is the increased volume and perhaps sharp change of sound fools the de-clicker into thinking it’s not a click. So I was thinking of how best to up my game, so to speak. From what I have noticed, adding bands doesn’t take a whole lot of more processing time, so maybe I should shoot it up a LOT, like up to 30? What do you think? Also, I’m wondering if other settings like dense click threshold, maximum click length, and widen repair intervals would yield any benefit if messed with (and if so, what direction!
). My clicks tend to be on the higher range, and tend to coincide with bursts of sound rather then between sounds.
Thanks Paul!
-Marcus
Marcus, have you tried Audacity version 2.1.0 yet?
The new spectral editing feature gives another means of taking a few remaining stubborn clicks out by hand.
But it needs some explaining and we don’t really have tutorial material yet.
Though you could watch this demonstration of the feature in action while it was still in prototype.