Up until a year or so ago the project was run by part-time volunteers, donations were welcomed and were used to fun the running of the online infrastructure, servers, bandwidth etc.
Latterly Muse, a commercial company, took over the running of Audacity. Accordingly when that happened it was no longer appropriate to accept or solicit donations - so the donation links on the Website were removed.
The project remains an open-source project and thus contributions from appropriately qualified volunteers (coding, documentation, user support) are welcomed for possible inclusion in the app.
You could easily answer this question yourself simply by trying it and see what happens. Does it sound better, worse, or no different after the second application?
I know nothing of audio. I just got a recording and ran the Desibilator in the default settings. There is still an issue with words that finish with ‘s’ that should be a ‘z’ sound, but there’s a sibilance. For instance, in the word “tatters” or “moves.” What can I try to fix it?
Is there an explanation somewhere of the different settings?
I’ve been using it for a number of years, but on my most recent project, (after the latest Audacity update of 3.1?), I’ve noticed that when I apply the effect to a longer audiobook chapter,(say 45 ~ 60 minutes), the process begins, but seems to time-out about 1/3 of the way through the progress bar and stops. I check the timeline, and the effect has not been applied at all.
I’m using the default settings for the effect.
As a test, I tried applying the effect to a shorter segment of the timeline, in this case, 15 minutes, and the effect seems to have been applied successfully.
Just to be sure, I de-activated, deleted, and re-installed the effect with the same results.
Any ideas on how I might be able to apply the effect to the entire timeline successfully again?
Thanks very much!
If anyone needs to contact the developer (Paul) of this plug-in, he is on the Audacity Dev discord channel https://discord.gg/N3XKxzTrq3
His handle on that channel is Paul Licameli#6950
(Note that the “Audacity Dev” discord channel is for developers and is not a general help channel)
Can someone help me install this declicker and deesser in my audacity? I followed every you tube video I can possibly watch. But none was helpful in installing this in my audacity folder plug in. I know that before it appears in the effects, it should appear in the audacity folder plugins. I am unable to do that.
When I try to download the de-clicker and the de-esser all I get is a download that doesn’t have any app to open it with. I’m on a Windows 10 version 22H2. I’m hoping you can help me with this. Thank you.
After reading through this entire thread over the past couple days, I decided to compile some experimentation guidance and explanation of the different parameters as I understand them. I’ve attached my own declicker settings for reference at the end, in case that’s useful to anyone.
Most important parameters for sensitivity and computation (source)
Frequency range and number of bands (default: 12 bands over 150 - 9600 Hz (2 bands per octave))
The main thing controlling both sensitivity and computation power needed is the number of bands. More bands = better click detection, but also more bands = slower. So, picking good values comes down to balancing those two factors. Under the default settings, the de-clicker processes 12 bands, from 150 Hz to 9600 Hz. This is one band per half-octave (150 * 2^6 = 9600, so 6 octaves). Low frequencies are more computationally expensive (see forum post), so if you don’t usually have clicks down that far, you can omit those frequencies from testing. For example, I mostly notice clicks that are higher pitched, so I set the range as 5,000 to 20,000 Hz. This is 2 octaves (5,000 * 2^2 = 20,000), so to have the same number of bands/octave as the default I would only need 4 bands. My current settings actually use 8 bands (4/octave) though, for increased sensitivity.
To calculate how many octaves your frequency range covers, if the high value of your frequency range is twice the low value, that’s one octave. If it’s four times (2^2), that’s two octaves. Eight (2^3) is three octaves, 16 (aka 2^4) is four octaves, etc.
Other parameters (grouped by topic, but no particular order)
Number of passes (default 2)
The De-Clicker enforces a minimum amount of time between clicks. If you increase the number of passes, this will allow you to catch clicks that were too near to another click to be detected the first time. This has diminishing returns and the author of the plugin says he never uses more than two. I use 1.
Step size (default 5 ms) / click length (default 2 steps)
Increasing the step size is a way to sacrifice some precision while making the plugin run faster. By looking at a spectrogram, I’ve found that I regularly have clicks that appear to be 15 ms or so, so I have step size set to 5 ms and click length to 3, so that 3*5=15 can actually capture those longer clicks.
Time between clicks (in steps, default 3) / dense click threshold (default -45 dB)
“Make [the time between clicks] too small, and more clicks will be identified, but there may be too much damage to voice when it drops in pitch. There is also a dB threshold (for the entire signal) below which closer clicks may be detected. This may help clean up crackles in de-voiced intervals and breaths. Use low-frequency rolloff effects [on the entire audio] first for better results!” (source)
Setting the time between clicks lower than the default may increase muffling, in particular for vocal fry that ends of sentences sometimes trail into (source).
Widen repair intervals (default 5 ms)
This specifies a “time period for cross-fading of repairs around each clicky interval. Make it too short and you introduce undesirable artifacts. This setting should not affect computation time significantly” (source). I have this set to 10 ms for some reason.
Bonus:
A neat trick to help experiment with new settings is to duplicate the audio you’re experimenting with settings on, and apply the de-clicker with “Isolate Changes” as the action instead of “Apply changes”. Then, look at that track in spectrogram mode. It will clearly show the clicks it detected for you to visually compare to the original track, and you can also listen to it directly to decide if it includes too much non-click audio. If you listen to it at the same time as the original track, it will also cancel out the clicks in the original audio, so that’s a way to preview what “Apply Changes” will sound like.
As ever when experimenting with an effect, manage presets is a great way to save and label your experiment parameters so that you can remember how to get back the one you actually liked declicker-settings.txt (316 Bytes)
I have a question about the attachment with an example of your settings.
I wonder whether I have a different version of this plugin, as some setting labels on mine are different to those in your document - which means that I had interpret your labels to mine, rather than copy label to label. I think I have done that correctly, but I’m curious why none of my labels has the word "control"in it?
and for example, is your label
control-step-dur-ms=“5”
the same as my
widen repair intervals at each end by (ms)
This is my version of the plugin - please note the different setting labels.
