Here The audio Files. First the original, then the RX version then the clicks only. Feel free to download the original and make a SL version for comparison. RX was as easy as plug it into the insert, done !
Original clip
RX Mouth De-Click
RX Clicks Only
I was afraid you would find a “good” example. You certainly did 
I could not find a setting in SL using the De-click module that had the same result so i didn’t bother to post an mp3. Please feel free to show your settings if you found a way to a similar result. Again, RX was for me the easiest: insert , set threshold, process, done.
Thanks for sharing these mouth clicks examples, I’ll certainly use them to fine-tune SL13 !
If you have others feel free to share.
Hi Robin, nice to finally meet you and thx for the fast reply. I’m glad these examples are helpful.
Most of the older SL video tutorials on steinberg’s youtube with his voice on it will be great/perfect for mouth noise removal testing because of :
Here is another example from his voice to use and test/refine the de-click / mouth noise functionality:
https://www.youtube.com/watch?v=iFtR2O2c3Dc
Thx again for this great tool and for listening. I’m looking forward for the new updates. Please let me know if anything new in this particular case happens, i’m glad to help !
Hi Robin,
I have a few more test files you can use for analyzing.
The first one is the most normal one and contains a wide range of typical mouth noises as well as clicky sounding consonants. It is many clips put together with a short fade inbetween.
male, spoken, large condenser mic, soft background music
male, spoken, shotgun mic
male, spoken, nasal mic
female, whisper, very strange close-up, but might be useful to analyze
Robin, audio examples aside, please consider what we call a “click” when editing, specially dialog.
Whenever an editor refers to “a click” it is 99% of the time a very clear transient spanning a considerable part of the spectrum. It literally looks like a vertical bar on the spectrogram. It sticks out like a sore thumb, there is no doubt about it when we see one.
As it is, the DeClick module finds all kinds of “specks” of energy scattered around in time and frequency and leaves those “bars” behind. This just screams that it simply does not “know” what a click - in a practical, useful sense - looks like.
If you could rename the current the DeClick module to “DeCrackle” and make a new one that is simply fine tuned to remove and heal only these very distinct transients, that would be an ABSOLUTELY FANTASTIC improvement already.
You already have all you need: Transient Selection and Heal. Just put them to work together.
By the way, if you can find a way to make the speech detection algorythms help the DeClick module preserve natural transients in consonants while removing “clicks”, you will blow every competitor out of the water.
I can not get good DeNoise results from Spectralayers 12 for say cleaning the amp noise from an electric guitar. RX Spectral DeNoise always does a much better job
what is your method for this process in SL? Run DeNoise module?
removing hum and buzz in spectrogram manually is one of the easiest things to do
I used to struggle with tools like Acon DeNoise and Sonnox Restore…the only thing I got to work well was Sonnox DeBuzzer loaded with the electric guitar PU factory preset which did return magic results.
I have only removed/ reduced electrical hum and buzz in location recordings with SL so far, but the results have been amazing with constant hum or buzz in SL.
The difficulty with vocal anomalies in singing (generally) is the need to differentiate between “normal” consonant transients vs unwanted anomalies such as lip smacks between words. Lip smack sounds and similar transient noises within words may be fine. You might not want to remove the high frequency transient at the beginning of a “P”, “T” or “B”, for example. Often, removal isn’t the best option - instead, you may prefer to simply attenuate the anomaly in a very specific way, instead of removing it. For this reason, I tend to manually edit anomalies on a per-incident basis in lead vocals (via spectral editing). I’ve never found an automated process which does the same job with appropriate discretion, although I do use automated de-clickers for backing vocals (which often have clusters of lip smacks when bussed). I often also manually adjust breathing sounds and shhh sounds on lead vocals. It’s partly also function of time vs cost. I don’t always have the option to do it manually, but when I have the time I prefer to use manual spectral editing, finessing each anomaly on a case by case basis.
Some engineers like to attenuate all consonants in backing vocals for some genres - often, pro backing vocalists are really good at attenuating consonants and esses/shhh sounds etc while singing, because they know it builds up and gets in the way of the lead vocal.
An AI model which does that on backing vocals would be cool i.e. aggressive attenuation of consonants, other transients and ess/shh sounds.
Yes I bring DeNoiser up on aguitar track, register the hum before the guitar begins and process it. The result is never as good as RX Spectral DeNoise for me. Is there some better method?
indeed, this is very well put and I do exactly the same: manual selection and edit…sometimes removal works on wanted transients and sometimes attenuation is the fix
that said, in most cases for me, that is after bouncing a Acon DeClick
manual selection 
One of the beauties of using SL (or generally spectral editing) for this is that you can simply select each harmonic very precisely and then attenuate or heal vertically until it sounds good for you. Results are much better than with spectral denoise, since there is no gating and with interpolation the overall energy at those frequencies is kept at a reasonable level.
Very well put.
The thing is that specially in the context of post-production (my main focus here) we more often then not have to deal with many clicks that are extraneous to the dialogue itself. I’d also say that in post, speech is much more forgiving regarding mouth clicks than in music. IMO there are many more opportunities for expressiveness in mouth noises in dialogue than in singing.
So with SL’s ability of processing only the selected areas and an improved DeClick module, I could select a buch of time during a vowel sound, or a breath, or whatever, and press a single button, instead of having to select each and every click individually (or the alternative, use Acon DeClick beforehand, as Tood put it).
Also that’s why I said that if Robin could make SL’s speech recognition work together with the theoretically new DeClick, perhaps we can even start selecting across consonant sounds without worrying about mangling speech.
I’m still using Spectral Layers 11, but we’ve conducted an in-depth hands-on comparison and workflow analysis. Once we get our hands on version 12, we’ll likely run a dedicated test with some truly challenging footage.
Love your work! Please do!