Pinpointing the mouth click.


though my mouth clicks are finally a lot less I still sometimes end up with a few and every now and then, during a word.

The de-clicker is fantastic but doesn’t always get them and I’ve become fairly good at finding and cutting them. I can sometimes tell just by seeing the waveform now.


I am currently stumped on one. I’ve already given up and shall re-record the word but it had me wondering whether there are any tools or tricks I’m unaware of that assist in this process.

I know how to zoom in and play at slower speeds.

Is there anything else please?


Clicks can sometimes show up most clearly in “Spectrogram” view, appearing as higher intensity (red or white) dots or lines:
Generally you would want to increase the “Window Size” (“Edit > Preferences > Spectrograms”) so as to get better frequency resolution - try it at about 1024 or 2048. If set too large the image become too blurred horizontally, and screen refresh can be rather slow.

Try it out first in places where you know exactly where the click is, then you will know what to look for in more difficult cases.

Thanks, Steve. :slight_smile:

This tutorial describes the process:


Thanks, Waxy. :slight_smile:

Woohoo and eureka!

The bane of my voice career thus far has met its match!

It’s mighty name is SPECTROGRAM!

Thank you again. :slight_smile:

And… If you dare… learn the Spectral Editing effects! Assign them to keystrokes (I use a, s, and d for multi, shelves, and parametric)! Then, when you see a click, drag a rectangle around it, and use one of the effects!

Greetings once more.

I’m finding the Spectogram very useful but sometimes I’m running into scenarios where it doesn’t clearly identify the click.

For those more experienced with this view, any suggestions when coming across something like this?

The green is my selected section where the click seems to be.

Thank you in advance.

Edit - Ah, I’ve found that one can select not just start and stop but low and high in the section. I was able to find it by moving the inside of my selection higher. It certainly doesn’t identify as clearly as many though, eh? :slight_smile:
Spectogram shot.jpg

Maybe you’re having trouble finding it because it’s so buried in the other sounds it doesn’t make any difference in the show? Koz

Aye, it is quite probable that the average listener wouldn’t even hear it.

But err, this is me. I’m a perfectionist when it comes to something I’ve created and put out into the world. :slight_smile:

I aim to keep my zero rejected by ACX score. :wink:


I aim to keep my zero rejected by ACX score.

ACX has said, multiple times, that nobody screens the works. Once the presentation makes it past the robot, it’s between you and the rights holder. Since there is an ACX failure of “overprocessing,” I suspect there is provision for spot checking the work or maybe comparing you to a spreadsheet listing “troublemakers,” (similar to the vetting provision on the forum). But past that, it’s just you being obsessive.


“A great work of art is never finished, only abandoned.”

Have you tried ctrl-P, then choose Spectrograms, then try other Preferences?

Without getting into the math, I can tell you I tend to like Blackman-Harris a little better than Hann, and I use 512 or 1024 for window size.

With a longer window, you get finer resolution along the vertical axis, but things get a little more smeared out horizontally. A mouth click is a brief thing you try to isolate. But my experience is that 1024 is not too much.

I find the subtle mouth clicks that survive the declicker often live somewhere around 4 to 5 kHz, and are a sort of little red pimple on the spectrogram – not very tall ones.

Options for spot treatment could be 1) declicker with the maximum number of bands (which would be slow applied to a whole track), or 2) drag a box around the click and try one of the Spectral Edit effects. I might do the more gentle Parametric EQ with -20 dB gain for one of those pimples inside a vowel sound.

More: the longer windows start to resolve the overtones in your vowel sounds as ridges oriented more or less horizontally (though with rises and falls). Sometimes the subtle click you hear shows up with that resolution, as something sticking up between those ridges. It will look like it doesn’t fit the pattern.

And my!, what long input/output meters you have, Grandma.

What word are you speaking there?

I have stared at enough spectrograms that I’m guessing it is something beginning with SH–

No joke, that is really my guess, because the sibilant sound at left seems to peak lower than an S normally does.

I don’t recall, Paul but you are probably right now that we know my new microphone has shown me I have that issue too. :frowning: