De-Ess Vocals: 10 Simple Steps to Remove Sibilance

De-Essing Vocals: The Complete Workflow

Sibilance is the harsh, piercing “s”, “sh”, and “t” sounds that make listeners reach for the volume knob. Here’s how to kill them cleanly without turning your vocalist into a mumbling mess.

Done right, de-essing is invisible. The listener never hears it working. They just stop flinching. We’ve mixed vocals where sibilance was so bad the “s” hits were clipping the bus by 3dB. After a proper de-essing pass, the same take sounded professional. That transformation takes under 20 minutes when you know the workflow.

Ten steps. Specific numbers. Let’s go.

Step 1, Find the Sibilant Frequency Range First

Don’t guess where the problem lives. Measure it.

Put a spectrum analyzer on your vocal channel (we use SPAN by Voxengo, which is free). Solo the vocal and loop a phrase with several hard “s” sounds. Watch where the energy spikes during those syllables.

Male vocals typically offend between 5-6 kHz. Female vocals typically push that problem zone up to 7-8 kHz. Rap vocals, because of close-mic delivery and faster syllable rates, can stack sibilance densely across 5-10 kHz. Classical vocals with more open mic placement usually land narrower and lower, around 4-5 kHz.

Write down your specific peak frequency. You’ll use it in every step that follows.

Step 2, Check Your Gain Staging Before Anything Else

Compression amplifies sibilance. That’s not a theory. It’s physics.

When a compressor reduces gain on transients, the ratio of sibilant energy to musical energy shifts. A vocal running into 4:1 compression with fast attack can sound twice as lissy after the compressor as it did before.

Keep compression ratios below 2:1 for anything sitting before your de-esser in the signal chain. If you’re already running heavier compression upstream, you’re creating extra work for yourself. Restructure the chain: de-esser first, compressor second. Gain stage your vocal to peak around -12dBFS on the loudest consonants. That gives you headroom before the processing stack.

Step 3, Set Up Your Dedicated De-Esser Plugin

Load your de-esser on the vocal channel. We’d reach for FabFilter Pro-DS as a first choice. It’s not cheap (check the manufacturer’s site for current pricing), but the wideband versus split-band modes and the visual feedback are genuinely worth it for the speed of dialling in.

If budget is tight, TDR Nova (free version) handles de-essing in dynamic EQ mode well. It’s slower to set up but the results are comparable for non-complex vocal parts.

Set the frequency detection to your measured peak frequency from Step 1. Start with a threshold where the gain reduction meter shows 6-9dB of reduction on sibilant hits only. If it’s catching everything, your threshold is too low.

Attack should be 0-1ms. You need to catch those consonants before they happen. Release between 50-100ms. Longer releases cause pumping.

Step 4, Choose Wideband or Splitband Mode

This choice matters more than most tutorials admit.

Wideband mode turns down the entire signal when a sibilant is detected. It’s transparent but can feel like micro-ducking when the vocal suddenly loses body during “s” hits.

Splitband (sometimes called “split” or “filter” mode) only turns down the frequency band where sibilance lives. This sounds more natural on most vocals because the chest and body of the voice stays constant. We use splitband by default and only switch to wideband if a vocal has inconsistent dynamics that make splitband artifacts noticeable.

For rap and hip-hop vocals with heavy low-mid proximity effect, splitband is the only option that works cleanly.

Step 5, Layer in a Dynamic EQ for Surgical Work

A de-esser handles the main offenders. A dynamic EQ handles everything else.

Load FabFilter Pro-Q 3 or TDR Nova after your de-esser in the chain. Set up two dynamic bands.

Band one: target 2.4-8 kHz, dynamic ratio 1.1:1, attack 0.1ms, release 250ms. Threshold set to trigger only on the loudest sibilant bursts. This band catches the broad upper-midrange harshness.

Band two: target 8 kHz and above, ratio 1.2:1, same attack and release. This catches the extreme air frequencies where “sh” sounds splash.

Together, these two bands catch what the de-esser misses. We’ve fixed vocals in a handful of minutes using this combo that a de-esser alone couldn’t tame.

Step 6, Automate Clip Gain on the Worst Offenders

Some sibilants are so extreme that no amount of plugin processing sounds clean. You need to physically lower them before they hit the processing chain.

Zoom in to a sibilant hit at the sample level. Use your DAW’s clip gain (not volume automation) to reduce just that consonant by 5-10dB before the plugins see it. This isn’t a last resort. Mixing engineers who skip this step are leaving quality on the table.

We once had a single “s” in a bridge that was 11dB louder than every other consonant. Nothing that didn’t destroy the rest of the vocal could fix it in the plugin chain. A 9dB clip gain reduction on that one syllable, and the de-esser handled the rest cleanly. Satisfying.

Step 7, Check How De-Essing Interacts With Your Reverb Send

Most people miss this one completely.

Sibilance sent to reverb returns as a harsh, washy “ssss” tail that smears across the mix. If you de-ess post-fader on a send, the wet reverb still gets the untreated signal. De-ess on the insert chain before the send. Your reverb tail gets the cleaned signal.

The difference is dramatic on longer reverb tails. We ran a vocal through a 2.4-second plate reverb both ways. With pre-fader de-essing, the “s” sounds created a 1.5-second sibilant smear in the reverb tail. With insert de-essing, gone.

Step 8, Handle Bus Compression Without Re-Introducing Harshness

Your vocal bus compressor will see whatever your de-esser outputs. If your de-esser threshold is set too low and it’s removing 12-15dB constantly, the bus compressor will hear a weirdly even signal and may over-compress.

Set your de-esser to touch only the genuine sibilant hits. 6-9dB of reduction during actual “s” sounds, near zero the rest of the time. Then your bus compressor handles the vocal like it should, as a normal dynamic signal.

At the mastering stage, a limiter can re-trigger sibilance issues if the vocal sits too loud in the master. Check your de-essing holds up by soloing the vocal at 0dB (true peak) into a limiter. If sibilance returns, you need to go back to your session and do more clip gain work.

Step 9, Reference at 79-85dBSPL to Catch What You're Missing

Low-level monitoring reveals sibilance problems that disappear when you’re listening loud. At high playback levels, our ears’ equal-loudness curves boost high frequencies naturally. You stop hearing sibilance because everything else competes.

Mix your vocal de-essing at 79-85dBSPL on a calibrated monitor setup. At this level, if sibilance is still present, it’s genuinely present. This is where we’ve caught nasty residual harshness that seemed fine at louder levels.

Don’t have a calibrated setup? Use a free SPL meter app on your phone. Point it at your listening position and adjust your monitor volume until you hit the target.

Step 10, A/B the Full Chain in Context

De-essing always sounds different in context versus solo.

Bypass your entire de-essing chain (de-esser, dynamic EQ, clip gain adjustments) and listen to the vocal in the full mix at the correct playback level. Then re-engage everything. You want the de-essing to feel invisible. The vocal should sound natural and smooth. If you notice the absence of sibilance in an obvious way, like the voice suddenly sounds lispy or hollow, you’ve gone too far.

The target isn’t zero sibilance. It’s controlled sibilance. A small amount of natural “s” air makes a vocal sound real. Kill it completely and it sounds ugly.

Pro Tips

Use iZotope’s AI-assisted detection when you’re stuck. Neutron’s and Nectar’s Unmask and Learn functions can identify the exact sibilant frequency for you in seconds. It’s underrated as a diagnostic tool even if you don’t use the processing it suggests.

Different genres need different de-essing aggression. Metal and hard rock vocals can tolerate, and sometimes need, more sibilant presence to cut through dense guitar mixes. Classical and jazz vocals need extremely light touch or listeners notice. Rap vocals usually need the most work because of close-mic placement and rapid delivery.

Mono check your sibilance. Sibilance is often panned center and can sound different in mono. Collapse your mix to mono and do a final sibilance check. It exposes harshness that stereo width was hiding.

Mid/side processing on the stereo bus can help. If sibilance is slipping through to the master, a narrowly targeted dynamic EQ on the mid channel at the stereo bus, set to 6-8 kHz, can catch what everything upstream missed. Use this as a safety net, not a primary fix.

Common Mistakes to Avoid

Over-de-essing until the vocalist sounds lispy. We hate this. It’s the most common error and the most obvious to listeners. If someone says your vocal sounds lispy or muffled, your de-esser threshold is too low or your frequency target is too broad. Back off until consonants still sound like consonants.

Using one de-esser to fix a problem that needs three tools. A standalone de-esser plugin won’t cleanly handle severe sibilance on its own. The layered approach (clip gain, dedicated de-esser, dynamic EQ) exists because each tool solves a different part of the problem. Using one tool too aggressively trying to compensate for missing the others sounds frustrating and artificial.

Forgetting to automate between sections. A vocalist’s proximity to the mic changes between verses, choruses, and adlibs. One de-esser setting rarely works across an entire song. Automate your threshold or use separate instances per section for results that actually hold up.

Setting a fast release and ignoring pumping artifacts. A release under 30ms on a de-esser causes audible pumping between sibilants. If you hear the vocal gain fluctuating weirdly, your release is too short. Start at 80ms and slow down from there until it disappears.