What is your de-essing strategy?

[cryophonik]

Let me start by saying this is NOT a name-your-favorite-de-esser-plugin thread. It’s in the Production forum because it’s meant to discuss techniques for minimizing and reducing sibilance, which also includes plugins, but that’s not the main point. I’m interested in hearing what you do during recording to minimize it, and how you process your tracks to reduce it.

I feel like I’m constantly fighting it. Most of the vocals I record are female singers I work with. I try having them sing with the mic off-axis, sometimes try different mics (although I don’t have a locker full of them, currently just 3 mics). In the past, I’ve owned outboard channel strips with de-essers, which usually helped a bit, but also sometimes clamped down too hard and would ruin a take. There are probably some tips I could give the singer, but I usually don’t think of it in the heat of the moment, and it seems like it would take their focus off their pitch, lyrics, and delivery.

In post, I usually use a combination of surgically automating the levels of the worst offenders to reduce them, then clean up the rest with either FabFilter Pro-DS or e2 Eiosis, my two go-tos. Recently, I’ve been using Melodyne’s sibilance tool with good results. But, I still listen to other people’s tracks and their S’s are so smooth and natural while mine still have this annoying piercing quality.

Anyway, I’d love to hear what other people are doing.

[swTB]

One way is listening to all of the music, not to the S's.
It never bothers me unless I zoom into it.
Extreme cases aside.

[jamcat]

I use a Stedman pop filter.
https://www.sweetwater.com/store/detail ... -gooseneck

If necessary I’ll use Melodyne to treat sibilance.

https://youtu.be/hJmGgcJVBQs?si=N2T3TmHW2e1f6jr3

If I want to actually have it sound like it was processed with an ‘80s hardware de-esser for a stylized production aesthetic, I’ll use the Lindell Audio model of the dbx 902.

[Tiles]

How does it sound without mastering?

[bmanic]

Nasty esses can really drive one crazy indeed. It's one of those things that takes me completely out of the mix, the technical mindset, the quickest. When I'm dealing with bad esses I need to constantly remind myself to take breaks in between. I shouldn't work on a vocal sound or the ess cleanup for more than 15 minutes at a time and then force myself to have a 30min break or work on something completely different (like the kick or bass) instead.

If you get stuck for 60min+ trying to solve the issues with an 'ess' then you've already lost and probably need to start over the next day. At least that's been my experience. It's also very important to always listen to it in context, not in solo.

This is why a discussion about this is indeed very valuable and an awesome idea to keep it here in the production techniques forum. Perhaps we could share some short problematic audio clips? The only issue is that the clips need to be long enough to contain a full melodic (or non melodic spoken word) phrase that makes sense. You can't just listen and attenuate esses in a vacuum. The processing can sometimes feel "good enough" if you listen to things in isolation, only to notice that it possibly doesn't work for the whole.

Another issue with esses is that they can sometimes be enhanced (or subdued!) by other instruments in the arrangement. If you get unlucky with a snare, shaker, hihat or any other "fizzy" instrument or percussion their phase may just perfectly align with the ess. And sometimes you get away with murder where the vocal in isolation has a really nasty ess but gets completely buried by the rest of the arrangement (full masking taking place).

So yeah, this is indeed a very complex matter.

As for recording, the off-axis mic technique can sometimes be a fix but sometimes it can be even worse! Why? Because going off-axis reduces the very top frequencies of the microphone which in turn means that the esses may stand out even more due to the energy above the esses not helping with the masking. Which is also true for EQing and de-essing. Sometimes you need to actually boost the air frequencies to get rid of the annoyance of an ess at 8 to 10kHz! Masking is such a weird phenomenon and the mystery here is the human ear-brain connection, not just the technical frequencies themselves. I'm sure many of you reading this have noticed that sometimes a really loud ess according to a spectrum analyzer can actually sound really pleasant and not at all annoying, whereas a really soft ess (amplitude wise) can still sound annoying as hell! Right? It's really weird how it all works.. and even after 30+ years of working with audio it's still a complete mystery to me how it really works.

I'll write another post with some concrete tips and tricks that I've used over the years but for now I feel like it's very important to acknowledge the complexity of this dilemma. It's a really deep and an annoyingly difficult subject and there are no one-magic-thing to fix it all. Unfortunately. Though I do have high hopes for AI processing in the future. A properly trained LLM may possibly be that magic bullet at some point but as far as I know, nobody has done that yet. It's such a delicate balancing act and esses sometimes contain the very essence of an emotional word or phrase so they can not be destroyed or attenuated with impunity.

[osiris]

High shelf filter cut at 8200. Usually on the backing tracks. My new favorite technique is one centered vocal and two clones on another channel both hard panned. A de-esser will help on really hissy stuff but High Shelf.

[donkey tugger]

Effects wise one unsung plug-in I rarely see mentioned is Airwindows 'DeEss'. There was some long-winded explanation as to how it works in a different way which I never understood, but I've found it very effective and transparent in most cases, and I've tried most of the others...

[Dirtgrain]

It's like pruning hedges: snip, snip here; snip, snip there; I just keep going until I've unintentionally reduced the vocal to a pathetic, Charlie-Brown-Christmas-tree-of-a-bush. Then I delete all plugins on the track, and start over. I'm joking--but there is a good amount of truth in it.
Seriously, I've been using Klevgrand's de-sser, Esspresso. I do wonder if the Fabfilter one or the Softube Weiss one or some other one are any better.

[osiris]

FL Patcher has a great one with a huge frequency range and a great Listen function that lets you hear what you're removing. Love it.

[imrae]

Step 0 is to aim for consistency in performance, step 1 is comping; if one take has a messy sibilant consonant, maybe prefer another take where it flows better. If no such take is available, then volume automation.

When the overall sibilance in performance is consistent, then we can start tinkering with EQ and dynamic processes.

[lfm]

I tried the technique with gating over to another track, and swap phase on that and tune in to desired effect with original on a bus.
- it will cancel out just those tops

It takes some experimenting, but a well mixed track is to live forever, so what is that time spent.....

[bmanic]

Okay, I promised a few tricks.. well here is one to start with, using FabFilter Pro-DS.

1.) Copy the offending vocal track to a duplicate track
2.) Insert Pro-DS on the duplicate track and set it up to do the de-essing
3.) Now turn on the "delta listen" function of Pro-DS
4.) Then flip the polarity of the whole duplicate track
5.) Then insert a linear phase parametric EQ and create a few tight peaks around the worst offending "whistling" on the nasty esses and boost them a lot (like +6 to +12dB)
6.) Use one or more wide Q nodes on the same linear phase EQ and cut some of the natural parts of the ess so that you mainly get to remove the annoying parts while keeping the wanted parts of the 'ess'. This helps to keep the 'ess' very natural and to avoid the dreaded lisping
7.) Control the overall amount of de-essing happening with the fader volume of the duplicate channel
8.) After this you can combine the two tracks into it's own "fixed" vocal bus and do all the rest of your processing here. You can easily automate the de-essing amount because it is now behind a channel fader (the duplicate track). Very handy and very quick to tweak. Note that if you increase the volume of the duplicate track too much you will not get de-essing but instead will be adding to the ess. Find the sweetspot.

Here is an audio example of the above idea in action:

Unprocessed very nasty ess

Processed ess and some post-tonal EQ and compression to shape the vocal

Google Drive links, 44.1kHz 16bit .wav files at about 230kb each.

Sometimes the ess is so bad that you just need to shape the whole aesthetic of the song to fit it. I sometimes end up making the vocal "lo-fi" on purpose, cutting all highs from 6 or 7kHz, forcing it into a different aesthetic. Naturally this means you have to mix the other tracks of the song to fit this new heavily shaped vocal and probably need to discuss it with the producer and artist first. Still, sometimes it's the only way to "save" a song that can't or won't be re-recorded.

[osiris]

I hear so many songs now where the vocal seems run through band filters and it's that effect I can't think of the name for ATM.
Not quite megaphone, but not a natural voice.

[legendCNCD]

Last time we had vocals was sooo long ago and that project did not see the light of the day in the end. I think we used pro-ds, but now I would check TDR Arbiter first, with Soothe2 I think.

[pvad]

I presume you are using pop-filters on the mic for recording? I know some engineers who use two pop-filters to help out with both pops and esses.