I edit a lot of dialogue and tend to remove unwanted clicks by hand since I’ve never found a Declicker that could be used indiscriminately without damaging consonants.
So I spend a good chunk of time with either transient selection + heal or marquee selection + heal, thousands of times a day.
So I was thinking, since the DeClicker module is severely lacking ( as if it simply doesn’t knowwhat an unwanted click in dialogue looks like, as I have stated multiple times in this forum), perhaps I (and others) could help to train the module in recognizing said unwanted clicks. As I’ve also said before, clearly transient selection + heal would make an EXCELLENT declicker if the program knows how to recognize one.
So if you want help and think that my logic makes sense, I am volunteering myself to use an SL build that somehow tracks what I do when using transient selection so it can watch me do it and go “Ooooohhh so THIS is what they’re talking about when they say Declick. Sure, I can do that!”
And I take this opportunity to remind you that we don’t really need it to be ultra-smart in recognizing consonants, if that takes a lot of development time. If we could just select broad sections of audio excluding what we want to preserve (consonants and contextual clicks), and press a module would be a HUGE advantage to selecting clicks individually as we need to do presently.
I’m well aware of RX Mouth De-click, and have used it for many years.
I think it is one of the best out there, specially on fiction work, but I still prefer to have more granularity when choosing what to keep, specially cause I work a lot with documentaries and there are a lot of desireable clicks in location audio. And it is faster to just do everything in spectral editing than processing a declicker on hundreds of audio segments through DOP.
Besides, Spectralayers needs to have a decent Declicker anyways, so my point stands.
Believe it or not, the unmixing levels feature works best for this (especially for vertical clicks). Such a shame that it cannot be previewed in-real-time.