Practical De-Essing Tips for the Mix

Hi, my name is Nick, and I have a problem. To fix the problem, I need to take certain steps, the first of which is to admit that I am powerless—specifically, I am powerless over my sensitivity to sibilance.

I’m an avid listener to radio, records, singles, and podcasts, so it can be a burden: If those esses haven’t been tamed, I wince with every offending syllable. And when I listen to an old project of mine and notice an unruly ess, I hang my head in shame, since the difference between balanced, silky vocals and insufferable sibilance is a tell, signifying an amateurish attempt rather than a professional product.

Perhaps you’re like me, and you suffer the sins of sibilance (also, alliteration and assonance). Or perhaps you will be like me: After prolonged exposure to chopping, comping, tuning, aligning, EQ’ing, compressing, and, above all, listening to vocals for hours on end, you’ll start to hear the offending ess. Once you hear it, you’ll never unhear it.

I have spent years combatting the dreaded ess, garnering a few tricks of the trade in my time. I am going to share some of them with you now, because the last step in making this problem better is to carry the message to other engineers, so that we all might practice the principles of de-essing in all our mixing affairs.

May Your First De-Esser Be a Manual De-Esser

There’s no way around it, at least not for me: The most natural-sounding de-esser consists of taking every ess you hear, separating its region from other syllables, and reducing its clip-gain by as many dB as you can get away with. If you use Pro Tools, the process is relatively easy—at least, compared to some other DAWs, where hotkeys aren’t as developed.

You could also split the esses out onto another track, but be careful in the way you edit between the two, because sometimes the audio will “pop” at the end of a sloppily cut phrase.

I use iZotope RX to handle clip-gain duties, because invariably, I’ll notice other things to fix before the mix (de-noising, de-plosive, weird EQ inconsistencies, some basic levelling, et cetera).

After a short time doing this, you’ll begin to recognize sibilance by sight; its waveform has a unique shape, almost like a football. You’ll also note that sibilance appears much denser than other kinds of waveforms. Soon you’ll be working relatively quickly, your eyes taking you from ess to ess.

Be forewarned though, this is not a failsafe against future de-essing. Depending on the processing you’ll employ, you might need to use EQ, dynamics-control, or de-essing plug-ins down the line. It will, however, sound more natural when you do.

Put a Wide-Band up Front and a Split-Band Later

Whether you’ve manually de-essed, you might find the following workflow helpful. I know that it certainly works for me.

I’ll tend to use two de-essers at different points of the chain, each one working minimally. One will always be the first processor. The other will come somewhere in the dynamics processing/EQ shuffle, depending on the vocal. I don’t use channel templates (I find them boring), though I can tell you the second de-esser is always tied to an EQ: Either I put it after an EQ to tame what I’ve just done, or before an EQ to safeguard what I might do. It depends on a host of factors (where the compression comes in—if at all; what the vocal sounds like, and so on). The important thing is the operational mode of the de-esser.

Most de-essers have wide-band and split-band modes. Wide-band mode affects the totality of the signal, so when its detection circuit hears an ess, it pulls all the audio down.

Split-band mode usually cuts the signal into two frequency bands, and from here, it gets fancy: The lower band could remain unaffected by the de-essing process, while the higher band receives treatment. In another variant, the lower band and the higher band remain untouched, and only the frequencies that you’re targeting receive processing.

I like to use wide-band before any other plug-in because, to my ears, it’s the closest you can come to manual de-essing without truly doing it. The second de-esser tends to act in split-band, since I’ve already sculpted the body of my sound in a manner most pleasing.

Working this way—with two processors operating gently in different points of the chain—I can achieve a more natural sound while taming my esses.

A Quick Word on Operation

There are a few ways to go about de-essing: You can audition the sidechain of the de-esser till you hear what you hate the most; you can use a frequency bump in an EQ to seek out troublesome spots, or slap on a frequency analyzer to visualize the auditioning process. From here, you can set a suitable range and tweak the threshold, or alternatively, tweak the threshold first, and work your range into the threshold. I tend to do the former, but certain plug-ins can work more naturally with the latter approach.

The Sylvester Effect

Whatever you do, train yourself to know when a de-esser sounds forced, because sometimes they can trick you—especially after a few hours in the mixing chair. An example: In researching whether the Eiosis E²Deeser was right for me, I came across a YouTube tutorial where user-error caused a horrible spitting sound to occur on particularly strong esses, as though the vocalist suddenly became Sylvester the Cat. Incorrect employment of any de-esser can easily cause this kind of behavior.

Now, if careful implementation still results in the occasional “Sylvester” effect, read on.

Automate

I don’t know about you, but I find the automation process annoying—necessary, but annoying (like voting). Once you’re locked into a choice, reversing it can be a cumbersome process (again, like voting). Still, there often comes a point when you’ll need to automate the de-esser. Perhaps you’ll want to change the band for a specific phrase. Perhaps you’ll need to raise the amount of reduction on an uncommonly strong sibilance, or lower it to avoid the “Sylvester” effect.

In these cases, automation is your friend. I can understand if latch mode feels intimidating for these moves, especially if you’ve got your de-esser set perfectly for nearly every other part of the song. So, go ahead and draw your automation in. You have my blessing.

Plug-in Specific Tricks

Hands down, the best plug-in I’ve come across for de-essing is the FabFilter Pro-DS. It’s the least destructive processor, to my ears, and it’s one of those rare plug-ins where the presets are truly fantastic. Other honorable mentions go to the Waves Renaissace DeEsser and the de-essing module in iZotope’s Nectar. I’ve gotten good results out of UAD’s VoxBox, but that’s a very specific all-around sound.

FabFilter’s Pro-MB multiband compressor can also come in handy as a de-esser, especially when the offending frequency triggers lower than you’d expect, say between 2.5 and 4 kHz. Sometimes I run into the following problem: The esses sound fine in the upper registers, but down low, they sound harsh when unprocessed, and cause an all-around lifeless sound when de-essed. Here Fabfilter Pro-MB is my go-to. Its GUI allows you to visualize the ess occurring higher up in the frequency band. From here, you can internally key the offending sibilance from this simply spotted—yet less harsh—ess. If done right, this can work transparently on solo vocals; I’ve used it when mixing episodes of Startalk Live, and in quite a few scenes of Hye Yun Park’s BKPI. You can use a sidechain input to use this trick with other plugs, but I find Pro-MB to be the most efficient.

Another workflow involves the stand-alone editor in iZotope RX 6 Advanced, which has a powerfully transparent de-esser. I get a lot of mileage out of the process in low, repeated doses. I tend to use it in Classic mode, as opposed to Spectral, passing one phrase through at a time, and threshold on a per-phrase basis. Then, if I want to control dynamics in a way I find more natural than compression, I run their Leveler module, and use its de-esser in low doses. Now I de-ess a third time, here using Spectral mode if a split-band approach is called for.

I’m knocking off very little each time—maybe half a dB, maybe at most 1 dB. By the end of the process, I have a vocal that doesn’t sound de-essed, and doesn’t irritate the ear.

You can leave out the leveler if you don’t need dynamics-taming yet and, conversely, you can manipulate the leveler so that it’s not doing much in the way of clip-gain boosting/attenuating, just affecting your esses. However, you will need to keep your intended RMS target in mind, or else the Leveler will bring the whole signal to a volume you didn’t intend.

I’ve got more tips up my sleeve, but these should get you started for now and, as always, feel free to leave a comment or question, below.