Just thinking about it…have a job atm and I have never been that happy with any deesser. I tried unmix components but it was very hit and miss
Does anyone else use SL for this and would you mind sharing workflow, tips etc?
Cheers
EDIT: The ideal would be splitting the esses onto a separate track and using a limiter as I have done experiments and that sounds amazing but now I have SL the labour should be a lot less right?
I’ve never been happy with automated de-essers.
Instead, I prefer to manually de-ess via spectral editing on a case by case basis, as each ess (or shhh or fffff) is different. But this is time consuming work and only suits people who are fussy and have the time. I use a different spectral editor (ie not SL) for this task, but you can do it with SL. I’m not claiming the tool I use is better than SL. Mostly, I adjust gain within a specific frequency area over a specified timeframe. The parameter choices vary depending on the needs for each edit. I’ve become fast at doing this due to having 20 years experience using my preferred spectral editor.
Yep I use to do that…but time is money as they say and I cant charge for the time…well actually Im just so over editing for the last 25 years.
That example of separate track worked the best…just hoping SL would do the heavy lifting
Thanks for your reply
Already Trier “Lisp” de-esser?
Works fine for me…
Have tried about 20 and given up…they fall apart when mixing into hybrid mix buss…suddenly you hear a lot of stuff that you never used to…Ill have a look though, even the latest ara stuff still doesnt solve it nicely…in this application
You could try unmixing vocals, then use one of the tools which separates transients from tones. Ess sounds should remain in the tones. Filter out everything below 2kHz, then remove all transients. Maybe you can then apply an automated de-esser to the result?
fwiw…melodyne has several methods that work quite well. My work is/has been all with singers and I use the “event by event” method since the “sss’s” occur quite differently depending on the “pitch range” and intensity of the word being sung…hth…/s~
AI models could be made to deal with various vocal anomalies, but the training data needs lots of ideal “before” and “after” files to do it properly. I suspect that’s been lacking in some of the existing “vocal cleaning” AI models, because the preparation of training files needs to be meticulous and edited by experts in the field. Lots of work!
I use Fab Filter Pro-DS
what the spectrogram can give you is specific freq areas to target
all that said, I wouldn’t call SLP fast
I would very much appreciate an in-depth discussion on this…more specifically, what are the SLP AI models trained on for something like Unmix Noisy Speech? I’ve asked before, yet we hear nothing
I’m assuming the models have been fed references that are generally accepted as “the best all round” processing?…because we always have remnants in the the unmixed noise from UNS and that is something I spend far too much time on!
I wouldn’t answer either.
“training models” as a thing, not verb, is a highly competitive area and becoming moreso imo.
I did do that (see my op) but some fragments remain in the tonal component eg hit and
I applied a basic rcomp to that selection which shows that it could work as the sibs were very smooth
Yes that was my first stop…the idea is great…even the Sib transition macro but it sounds noticeable when pushed even slightly
For sure there are many tools even sonible, fan, waves silk…prob is I have spent a fair bit of time and they don’t work
In this application
Eg Jack Johnson, all at once, the clarity and top lift by BG makes the whole thing an immersive space…but I know for a fact part of it was over treating Sib so that by the time it was finished…the vox was smooth and intact
Context in point
I wouldn’t say SL is the ideal tool for de-essing.
The de-ess functions in Melodyne is faster and more powerful. Easy to handle.
As for plugins there’s a similar concept in apulSoft’s SplitS, which lets you easily set the levels of sibilants quite transparently. There’s a 10 days demo.
My personal favorite is the SPL de-esser (and dual), that uses phase inversions of the sibilants to soften the S-sounds.
I’m no fan of compressor based de-essers.
I’ve never found a deessing processor sound as good as manual gain. But time is money, I know …
Thanks Ill check that out.
I was inspired by a synchro arts pitch tool that completely split the sibs etc and you could treat it separately…that was my hope for SL
Thats funny..back in SX3 days thats how I did it by painfully extracting the s track but it was very transparent.
Yes I thought that was going to be the fix as well but I got talking to them and they only handled the transition…which like I said is ok for minimal work. Dont really get the issue with ribbon mics but most stuff I get has some level of issue…just the nature of it all.
Thanks @Marshall for the suggestions…SplitS sounds good.
I use Melodyne but have good experiences with Sonible Smart:DeEss
Sorry if mentioned before, could not find it. 
Best, Pelle
Have you tried Smart:DeEss from within SL? I have (using Studio One) and it doesn’t seem to work. Not that it is important, but
Hi Henrik.
No, not from SL. (Pretty new to SL and did not know it’s possible to use other plugins within SL.)
I have just used Smart:DeEss as a plugin in Studio One and Cubase, but that is working in realtime.
To de-ess manually melodyne has been great for me.
Best, Pelle