The No.1 Website for Pro Audio
Airwindows DeBess: Mac/Windows/Linux AU/VST
Old 28th October 2019
  #1
Airwindows
 
chrisj's Avatar
Software Airwindows DeBess: Mac/Windows/Linux AU/VST



TL;DW: DeBess is an improved DeEss, with perfect rejection of non-ess audio.

DeBess.zip(355k)

This should be fun! Sometimes it's worth keeping a weird codger like me around meet DeBess.

Named because, for some of you at least, it is at last The Best De Ess. Period. Ever.

…assuming certain conditions.

DeBess is an extension of my former DeEss, which itself was the high point of several earlier attempts at a special de-esser with an unusual algorithm for finding specifically ess content and rejecting anything else in that frequency range, no matter how many overtones it had. DeEss used a set of sample comparisons to try and find esses, and was very successful at this… except some folks had trouble getting it to engage, and others needed it to be more perfect at rejecting even the faintest softening of other content.

DeBess does this by extending the sample comparison window a LOT. In fact, it's now a slider! You can set it to be even blurrier than DeEss if you like… or barely crack it open to replicate the original DeEss… or crank it up for high isolation de-essing. If you are recording on prosumer equipment or using moving-coil microphones, you might not get enough change between samples to engage DeBess. Same if you're using high sample rates and your mics do NOT extend right up as far as the sampling lets you: DeBess is not for taking stage mics and making them lisp. It is very distinctly for taking the most high-end of vocal tracks and de-essing only the ess sounds out of them, with zero cost to anything else. Whatever you're using, if your esses are blowing out the highest treble (which is exactly what you need a de-esser for) then it ought to work for you.

If you're using high sample rate and struggling to get DeBess action and you're going to be treble boosting for that ultra-bright voice sound, try brightening BEFORE DeBess and you'll probably be able to get what you need. It wants very bright esses to work with, so it can duck and darken them. Use the filter control to shape a better EQ on your esses, rather than just trying to duck 'em.

Airwindows is supported by Patreon, so if you can afford to buy this for $50 like it was a commercial plugin, and you'd want to, then go throw $50 a year at the Patreon instead. If I've written another good plugin or done something else useful a year from now, you can keep it going
Old 28th October 2019
  #2
Lives for gear
 
JulenJVM's Avatar
Yes! Finally, and right on time. Will be trying this in a couple of days on a pretty sybilant female vox track, will report soon!
Old 28th October 2019
  #3
If I were to pay $50 for a plugin, this would be the one. Increasing my Patronage, from $3/mo to $4/mo.

Great freakin' job.

Ultimately Spectacular!

Inspired me to write a book-post:

Isolation of removed ess'es material allows for the option of rendering them to a new track, with something like... say, harmonics plugins... then you could generate harmonics of the esses... which would allow you to be able to possibly get away with removing even more of the actual ess's!!!

I'm going to be inserting this into my mastering chain. I'll be probably also splitting into mid-side processing mode & then chose how the center (or sum) frequencies get DeEss'd... and DeEss the sides (or difference) frequencies at a different attenuation amount (pre and post gain adjustments).

I'm guessing that Spiral or UnBox would be easily used in a way that creates harmonics from the removed ess'es signal, and then render the difference between the clean removed esses & the dirty (post-DeBess) signal... to generate audio that is ONLY the added harmonics... and then mix that audio into the clean track with 2 channels... perhaps summing with Console6, or could channel 7 be useful for that?

I know that you're against multiband processing... but I'd be interested in finding out what it will sound like, if I find the bad Ess frequency that I want to get out (the center of the band that I want the most amount of esses removed... or most disgusting "ess" frequency point)... and then split it into two bands, or split it at just below or above the center (most nasty ess sounding) frequency point; and then process the Esses in the lower band one way & the esses in the upper band, another way... you could even split it into 3 bands, so that one small band is around the most nasty area, and the surrounding bands are including some of the esses that you want removed... & process all 3 bands at different intensities & different sharpness, etc.

I suppose that I could isolate the esses sound, then use a frequency analyzer to see where it is the loudest... & then use an EQ plugin that has solo'ing & dragging of bell shaped band, for listening to where it sounds the most crappy.... blind A/B it with and without the effect; and then A/B test it in multiband split results vs. broadband processing results. Could blind A/B/C test results with 2 band split, 3 band split, and broadband... to find out if the multi-band processing causes too much damage... since, as you've explained before, multiband processing introduces aliasing issues. It would be nice if the methodology described above (with use of harmonics) could be introduced into the 2 or 3 band split method... so that you could process each band of the ess'ing material with different types of even or even+odd types of harmonics (generally with some sort of a roll off at the highs, would work probably work best).

Could add odd, even, or even plus odd harmonics to different bands of where esses are getting handled. This way, even ones would make it smoother while also deessing, odd harmonics would add some sharpness or brightness while also de'ess'ing, and even plus odd harmonics while also dessing would do a combination of the two. What I'd want to do is pick the most unpleasant esses to remove, make a band around it.. and use even harmonics there, then on the surrounding bands, where I might want to add brightness, use odd or even+odd harmonics... and then, after creating harmonic material to sum with the original audio... go back to the original audio (with the same 3 band split), turn off the harmonics plugins, and then do final tiny adjustments to the DeEsser, before summing (such as increasing or decreasing the strength of how much ess material is removed).

I could also split into mid/side processing, on each channel that recieves the 2 or 3 band split... to pick whether the ess'ing of each section needs to be effecting the stereo image... since the esses may be in the center or center and sides, depending on how the material was mic'd (or some instruments cause an ess'ing soind). Or, split into mid/side mode, then split into 2 or 3 bands (but I prefer the aforementioned method).

The only problem is... its a pain in the butt to build all of that in the DAW, and it creates clutter & more probability of over-processing.


After that, I'll use post-dynamic EQ. Before using it, I'll run Ozone 9 dynamic EQ and tdr nova to see what spots are being dipped, around the ess'ing area... that'll help give me a guide, as to where I might want to center my middle band of a 3 band split would probably be best... & then listen by ear, to actually pick where it should go.

If you add harmonics to post/receive channels, then there shouldn't be any alias'ing... but it will add material to the audio, when you add the multiband split channels back to decoded stereo. That'll cause Ozone 9 and TDR Nova to find different areas of the audio, to do dynamic eq cuts.

Might want to use even harmonics and make the most loud (2-5 kHz) area of any ess'es done so that they're smoothed with even harmonics, and then 5-6 k with odd or odd+even, and 6-11 kHz with either even or even + odd (or perhaps with no harmonics, to preserve the highs).

Any thoughts, by others?


Possibly, do a tad of dynamic eq before de-essing in multiband mode... then a tad bit more dynamic eq, then a post-DeBess stage, in stereo...

After trying all of that, I'll most likely end up A/B'ing it with broadband results and deciding that broadband is the way to go (mostly to save time, but also because of the complications of multiband processing... perhaps just the mid/side processing is useful for de-essing stuff).
Old 29th October 2019
  #4
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by theMuzzl3 View Post
Isolation of removed ess'es material allows for the option of rendering them to a new track, with something like... say, harmonics plugins... then you could generate harmonics of the esses... which would allow you to be able to possibly get away with removing even more of the actual ess's!!!
Um, except you can't generate harmonics of esses.
Old 29th October 2019
  #5
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by chrisj View Post
Um, except you can't generate harmonics of esses.
Hey Chris,

I´m trying the plugin on 88.2 kHz session, and I´m struggling to understand what the controls do. I also find it difficult to make it work on more than just the stronger esses.

It´s a bit fiddly, but once you get some correct settings, it removes beautifully the high energy of the strong esses, in a very natural manner, with little time artifacts, and leaving the rest of the spectrum untouched. But I can´t get it to trigger with softer esses that still cause sybillance problems.

This is what I´m doing: I´ve looped a passage of a track that has a couple of very prominent S sounds, and I´m monitoring the difference with the Sense slider to the right, listening to what the plugin is removing, so I can tune the settings for optimal removal.

Please tell me if I understand what the controls do. I´m using compressor controls as a comparison in order to help put my thoughts into words:

*Intense slider: seems to work like a wet/dry knob, I don´t hear the plugin doing much for values under 0.8 (max=1).

*Sharp: seems like a sidechain filter to limit the spectrum the plugin is acting on. It looks like you have tuned it very wisely, you can really hunt down the esses, and the esses only with this control.

*Depth: seems like a threshold control that sets how much the plugin removes, but it seems reversed - the plugin removes much more sybillance with minimal controls. With Depth=1, I get no removal at all.

*Filter: seems to somehow mitigate the timing artifacts that are apparent if the depth is set too low. If set to high values, there´s very little sybillance being removed.

The things is that, once I´ve found good settings for this clip with loud esses, I move onto other passages with softer, but sybillant esses, and the plugin doesn´t get triggered.

What can I do to get settings that would work in general for the whole track, or do you recommend to process loud and soft esses separately? Could you please shed some light on the function of the controls above?

Thank you!
Old 29th October 2019
  #6
Airwindows
 
chrisj's Avatar
To DeBess there's no such thing as a 'softer ess that is too sibilant'. It ONLY cares about taking out esses that are big volume jumps in the ultra-highs. Intense slider is 'turn it up until it engages enough', so you'd be wanting to crank it. Sharp is the window for how it calculates, and more is a longer window that will really isolate esses. Depth is how much ess is left, so you can turn the ess down to zero or leave some of it; Filter is lowpass filtering the esses instead of simply gating them so you can use it to retain audio when the ess is being ducked.

Again, there's no such thing as an 'ess that is too sibilant but not loud enough': there is no such thing as a nonsibilant S. It's all about whacking down the loud esses in such a way that it sounds like nothing was changed at all. I'm glad it's still working at 88.2k as that was one of my concerns… are you treble boosting at all? If you are, you should do that BEFORE DeBess and that will help it engage. I'm kind of trying to grapple with the concept of needing to isolate everything that is an ess without regard to level and alter them: I feel like DeBess is VERY much of the opinion that some esses (and zees, and effs etc) should not be touched and other ones are simply way too loud. It's the way too loud ones that concern me.

Again, if you're needing to affect all esses (what about thhhhs? fffffs?) to de-treble them, I'd need to know more about what you're cutting and what you're keeping. And if this is because you're brightening stuff, the obvious answer is that it should go after the brightening and that'll help it engage.
Old 30th October 2019
  #7
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by chrisj View Post
To DeBess there's no such thing as a 'softer ess that is too sibilant'. It ONLY cares about taking out esses that are big volume jumps in the ultra-highs. Intense slider is 'turn it up until it engages enough', so you'd be wanting to crank it. Sharp is the window for how it calculates, and more is a longer window that will really isolate esses. Depth is how much ess is left, so you can turn the ess down to zero or leave some of it; Filter is lowpass filtering the esses instead of simply gating them so you can use it to retain audio when the ess is being ducked.

Again, there's no such thing as an 'ess that is too sibilant but not loud enough': there is no such thing as a nonsibilant S. It's all about whacking down the loud esses in such a way that it sounds like nothing was changed at all. I'm glad it's still working at 88.2k as that was one of my concerns… are you treble boosting at all? If you are, you should do that BEFORE DeBess and that will help it engage. I'm kind of trying to grapple with the concept of needing to isolate everything that is an ess without regard to level and alter them: I feel like DeBess is VERY much of the opinion that some esses (and zees, and effs etc) should not be touched and other ones are simply way too loud. It's the way too loud ones that concern me.

Again, if you're needing to affect all esses (what about thhhhs? fffffs?) to de-treble them, I'd need to know more about what you're cutting and what you're keeping. And if this is because you're brightening stuff, the obvious answer is that it should go after the brightening and that'll help it engage.
Hi Chris,

Sometimes I find it hard to find the right words to express my audio thoughts, so here´s an example.

I have attached some clips of what I´m cutting that contain what I call different levels of sybillance, or different volume jumps at the ultra-high spectrum. I´m struggling to find settings that work more or less for the whole track - I hope this demonstrates it.

Each file contains some short clips: Raw Audio (x2) + Audio Processed by DeBess (x2) + Difference between raw and processed (x2) - then comes the next clip.

Debess wav 1: I found some settings that more or less work for the 1st clip (I did it quickly to make this test), but cause time artifacts on the 2nd clip, and lisp on the 3rd
Debess wav2: these are softer sybillant clips, but that will likely cause problems once I start compressing the track and the volume gets brought up. DeBess does not engage at all with the settings above.

DropBox Link Sorry about the DropBox link, it looks like you can´t attach short 88.2kHz 24 bit wavs to posts on Gearslutz...in 2019....

FYI, the vocal brightness was added on the way in, I don´t think I will be boosting the highs much more on this track, but we´ll see at the mixing stage. It was recorded with an Audio Technica 2050, so maybe that´s part of the problem. Maybe it´s this particular track where ANY de-esser will have a hard time - I need to test this more on other material.

But right now, how I see DeBess in my workflow, is being the 1st de-esser on the chain, tackling only the most prominent issues, making it easier for other plugins down the line to work on finer issues. In the end, I always end up bringing down the esses manually because no de-esser does a perfect job, there´s always something. Or maybe I´m just clumsy and I don´t know what I´m doing...

Anyway, I appreciate your thoughts on this, I want more than you for this to be the best de-esser in the world! lol I hate sybillance when I´m mixing, it drives me nuts and I can´t stop untill I have sorted it out completely. I need an ultimate weapon that erradicates sybillance from this world lol
Old 30th October 2019
  #8
Airwindows
 
chrisj's Avatar
Well, you can do that if you want the audio to sound as close to untouched as possible. I think some people will want that: their goal would be to make it seem like they are NOT removing sibilance, yet it still doesn't become a problem. For your second example I think they'd go 'no, that's exactly how it should be'. Remember, esses ARE S. It's a bright white-noisy sound but rather highpassed: there is no way to change it into another sound short of further filtering.

I can do that with DeBess, even for your case I think, but you might end up using more than one instance. I can see using it as the first thing in the chain and by all means carry on with that… but then after you've compressed and brightened and all the things you want to do in mixing, consider putting a second DeBess on after all that, once you've got the vocal track you want.

You might be needing more thorough de-essing because you're bringing up what would not be objectionable esses… and by the time your chain is done, they've turned INTO objectionable esses. If that's the case, then DeBess will then be able to handle them better: they'll be louder and brighter and it'll be the same process.

I see that with the new sharpness control it's able to isolate the ess really well. That makes me happy

If you get lisp you need to continue engaging DeBess the way you were, but set the depth and filter until it SOUNDS right. There's not really any point where it's supposed to work with depth and filter at 0, that's just full power processing. It's probably filter you want to be using: that will roll off only the highs of the ess, if set correctly, and I think that's what you want. If DeBess is engaging or over-engaging, you modulate its maximum effect with depth and filter
Old 30th October 2019
  #9
Lives for gear
 

Looking forward to testing this on 2-trk Masters.
Old 31st October 2019
  #10
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by RJHollins View Post
Looking forward to testing this on 2-trk Masters.
Be extra sure to use the filter on 2-track masters! You surely don't want to gate the whole entire track! Also, pay attention to what the sense circuit is doing. I found that on mixed music, there were percussive sounds that were being affected over a really wide range.

Since this is not an isolation filter, all it can do is duck or filter the whole entire track, and getting isolation on just esses will become impossible. You can trigger on the ess only, just about, but if you're working on a full track with it, once you've triggered what can you possibly do? It's not a notch filter, it's either a complete gate or a lowpass filter. So set it up wisely and don't try to do too much: de-essing on 2-track masters is a dangerous game in the best of circumstances.
Old 31st October 2019
  #11
Lives for gear
 

Quote:
Originally Posted by chrisj View Post
Be extra sure to use the filter on 2-track masters! You surely don't want to gate the whole entire track! Also, pay attention to what the sense circuit is doing. I found that on mixed music, there were percussive sounds that were being affected over a really wide range.

Since this is not an isolation filter, all it can do is duck or filter the whole entire track, and getting isolation on just esses will become impossible. You can trigger on the ess only, just about, but if you're working on a full track with it, once you've triggered what can you possibly do? It's not a notch filter, it's either a complete gate or a lowpass filter. So set it up wisely and don't try to do too much: de-essing on 2-track masters is a dangerous game in the best of circumstances.
Thanks for the DeBess insights, chrisj.

With Mastering, I'm always on the lookout for De'Essers that deal with full mix situations.

I do have a couple plugins that look to address this ... but the choices are limited. When all else fail ... iZotope RX Spectral Editor for manual corrections ... very time consuming.
Old 31st October 2019
  #12
Gear Addict
 

I think this is a brilliant plugin! A little fiddly but that a side its one of the best de-essers I've heard.
Old 2 weeks ago
  #13
Quote:
Originally Posted by chrisj View Post
Um, except you can't generate harmonics of esses.
Oh yeah... duh! Lol.

Sometimes I am so dumb, that its entertainment!
Old 2 weeks ago
  #14
Here for the gear
 

Errors?

Hi Chris,

I am not a programmer, but, to the level of my knowledge of all things computer, I have a habit of occasionally running applications under Terminal or GDB (sometimes with a breakpoint in malloc_error_break set) to check for all sorts of errors and warnings that might pop up.

Just FYI, this is what your plugs produce in several hosts that I've tested:

Quote:
Getting metadata for AU: Airwindows: BitShiftGain
Reading symbols for shared libraries warning: Could not find object file "/Developer/usr/bin/../lib/gcc/i686-apple-darwin10/4.2.1/crt3.o" - no debug information available for "/var/tmp/gcc/gcc-5666.3~6/src/gcc/config/darwin-crt3.c".
. done

Getting metadata for AU: Airwindows: DitherFloat
Reading symbols for shared libraries warning: Could not find object file "/Developer/usr/bin/../lib/gcc/i686-apple-darwin10/4.2.1/crt3.o" - no debug information available for "/var/tmp/gcc/gcc-5666.3~6/src/gcc/config/darwin-crt3.c".
. done

Getting metadata for AU: Airwindows: DeBess
Reading symbols for shared libraries warning: Could not find object file "/Developer/usr/bin/../lib/gcc/i686-apple-darwin10/4.2.1/crt3.o" - no debug information available for "/var/tmp/gcc/gcc-5666.3~6/src/gcc/config/darwin-crt3.c".
. done
Is this a normal behaviour?

Last edited by ssp3; 2 weeks ago at 09:47 AM..
Old 2 weeks ago
  #15
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by ssp3 View Post
Hi Chris,

I am not a programmer, but, to the level of my knowledge of all things computer, I have a habit of occasionally running applications under Terminal or GDB (sometimes with a breakpoint in malloc_error_break set) to check for all sorts of errors and warnings that might pop up.

Just FYI, this is what your plugs produce in several hosts that I've tested:



Is this a normal behaviour?
I have no idea. "lib/gcc/i686-apple-darwin10/4.2.1/crt3.o"? Why would I have that? It sounds like part of the compiler, and the compiler does compile the plugins so as they work. You may be over my head here. I just make pretty noises
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump