De-Esser

A de-esser is a frequency-dependent compressor that reduces sibilance: the harsh, hissy energy of consonants such as s, sh, z and ch in a vocal. It listens only to the high-frequency range where sibilance lives and turns it down for a split second each time it gets too loud.

Used well, it makes a bright vocal comfortable to listen to without making it dull. Used badly, it gives the singer a lisp.

What is a de-esser?

Sibilant consonants are made by forcing air through a narrow gap between the tongue and the teeth. The turbulence creates a burst of broadband, high-frequency noise that can be much louder in the upper range than the vowels around it. On a recording, especially after compression and a treble boost, those bursts can become piercing.

A de-esser solves this automatically. Instead of you turning the vocal down by hand on every s, it detects the sibilant energy and ducks it for a split second, then lets the vocal return to normal. The rest of the performance, including the natural brightness of vowels, is left alone.

How it is defined

The Wikipedia entry on de-essing describes it as any technique intended to reduce or eliminate the excessive prominence of sibilant consonants. Antares defines a de-esser as a frequency-specific compressor that reduces sibilance in vocal recordings, and Splice describes it as a frequency-dependent compressor that detects the high frequencies where sibilance occurs and attenuates them once they exceed a threshold.

The name is literal: de- for removal, ess for the letter S. Unlike many audio terms, it has only one meaning.

A short history of de-essing

De-essing grew out of the physical limits of recording and playback media, which struggled with dense high-frequency energy long before anyone worried about harsh vocals on earbuds.

• 1939, film sound: the practice is traced to Warner Bros., where high-frequency limiting protected fragile film soundtracks (Wikipedia).
• 1960s, vinyl cutting: treble limiting was used when cutting records, because sharp high-frequency peaks are hard to cut into a groove and hard for a stylus to track.
• 1970s, dedicated units: Orban’s 516EC, designed for the recording and film industries, was one of the first dedicated de-essers.
• 1980, dbx 902: the dbx 902 compared the high-frequency content with the full-band signal instead of reacting to an absolute level, and worked over a wide enough range that it could almost be set and forgotten (AudioTechnology). It became a studio and broadcast standard.
• 1990s onward, digital and alternative designs: digital mastering processors such as the Weiss DS1 brought precise split-band de-essing to mastering, and SPL’s de-essers used a phase-cancellation approach instead of a conventional gain-reduction stage.
• 2010s and 2020s, plugins: software such as FabFilter Pro-DS added lookahead and intelligent detection modes, and resonance suppressors such as oeksound soothe (soothe3 was released in May 2026) and sonible smart:deess took an adaptive, spectral approach.

How a de-esser works

The sidechain does the listening

Inside a de-esser the signal is split into two paths. The audio path is what you hear. The sidechain (detection) path is only used for decisions: it runs through a high-pass or band-pass filter tuned to the sibilant range, and a level detector watches that filtered signal. When the filtered level crosses the threshold, the de-esser reduces gain on the audio path. When it falls back, the gain returns.

Threshold versus relative detection

Most de-essers use a threshold: gain reduction starts when the sibilant band gets louder than a set level. The weakness is obvious if a singer moves back from the mic or sings softly: the whole signal drops, sibilance included, and a fixed threshold may stop catching it even though it still sounds harsh relative to the voice.

The alternative is relative (level-independent) detection, used by the dbx 902 and by some modern plugins and dialogue tools. The de-esser compares the high-frequency energy with the energy of the whole signal and reacts when the ratio is too high. A whisper and a belt are treated the same way, which is why this design is popular for speech and dynamic performances.

Timing and amount

Sibilants are short and sharp, so de-essers react very quickly: attack times are typically around a millisecond or less, and many plugins use lookahead so they can start reducing gain just before the s arrives. Release must also be fast, often in the tens of milliseconds, so the vowel that follows is not pulled down. For a natural result, a de-esser usually works with a few decibels of reduction, often 3 to 6 dB on each sibilant; heavier reduction is possible but quickly becomes audible.

Where sibilance sits

Acoustic phonetics explains why a single fixed frequency often misses some esses. The /s/ sound has its spectral peak high up, roughly between 4 and 8 kHz, while /ʃ/ (sh) peaks lower, roughly between 2.5 and 5 kHz, and the exact values depend on the speaker. As a rule of thumb, male voices tend to have their harshest sibilance lower (around 3–6 kHz) than female voices (around 5–10 kHz). Treat these as places to start searching, not settings.

Types of de-esser

Type	How it reduces sibilance	Trade-off
Wideband	Detects high frequencies but turns down the whole signal	Simple and phase-coherent, but the body of the vocal dips on every s; can sound like pumping if pushed.
Split-band	Splits the signal and reduces only the high band	More transparent on vocals in a mix, because the low and mid content is untouched.
Dynamic EQ	A bell or shelf filter cuts only when triggered	Very precise, with adjustable width; no crossover filters, and no processing until it is triggered.
Phase cancellation	Adds an inverted copy of the detected sibilant band back to the signal	Avoids conventional gain-reduction artefacts; associated mainly with SPL designs.
Spectral / resonance suppression	Analyses the spectrum and applies many small, fast cuts where harshness appears	Very transparent and adaptive, but heavier on CPU and can add latency.

How to use a de-esser

Why vocals get sibilant in the mix

A raw vocal is often less sibilant than the finished mix. Two common mixing moves make it worse: compression, which brings up quieter details (esses included) relative to the vowels, and a presence or air boost, typically somewhere between 5 and 12 kHz, which lifts exactly the range where sibilance lives. As mixer Manon Grandjean put it in Sound On Sound, “the more brightness you add, the more the ‘s’ sounds are being emphasised” (Inside Track: Dave ‘Funky Friday’, June 2019).

Setting it up step by step

1. Find the frequency. Use the de-esser’s listen or monitor mode to hear only the sidechain signal, and sweep the frequency until you hear mostly esses and very little of the voice.
2. Choose split-band for vocals in a busy mix and wideband for gentle work on exposed voices or dialogue, then compare by ear.
3. Lower the threshold until only the sibilant peaks trigger gain reduction, not every bright syllable.
4. Aim for a few decibels of reduction, watching the meter and listening in context with the full mix.
5. Check for a lisp: if an s starts sounding like th, back off the threshold or range.
6. Bypass and compare at matched level. The vocal should sound smoother, not duller.

Where it goes in the vocal chain

There is no single correct position, and professionals use both of these approaches:

• Before compression: taming loud sibilant peaks first stops them from triggering the compressor, which then reacts more smoothly to the body of the vocal.
• After compression and EQ: because compression and treble boosts are what make sibilance harsh, a de-esser at the end of the chain catches exactly what the listener hears.

Many engineers do both: a light de-esser early and another gentle one late, each doing a little rather than one doing a lot. Grandjean, for example, used an SPL de-esser in her recording chain just after the mic preamp and another on an effects return in the mix.

Clip gain first

For very sibilant takes, many editors lower the worst esses manually with clip gain (a few decibels on each offending consonant) before any plugin. The de-esser then has less to do and sounds more natural. Some DAWs and tools can automate this kind of level-based de-essing directly on the clip.

De-ess your reverb and delay sends

Sibilance is especially ugly in effects: a single s sent into a bright reverb or delay becomes a hissy splash or a string of repeating esses. Placing a de-esser on the send or at the start of the effect return, before the reverb or delay, keeps the dry vocal crisp while the effects stay smooth.

Beyond vocals

De-essers are also useful on other bright sources: harsh acoustic guitar strings, cymbals in overheads, or a spoken-word track. On a full mix, a gentle split-band or dynamic EQ-style de-esser can tame a harsh vocal in mastering when no multitrack is available.

De-essing across genres and media

• Pop, hip-hop, trap and EDM: vocals are heavily compressed and mixed bright, so de-essing is usually more aggressive and often split across several stages, plus manual clip-gain edits.
• Jazz, classical, folk and acoustic: mic choice and placement do most of the work; any de-essing is kept gentle.
• Film, TV, podcasts and voiceover: levels vary a lot as speakers move, so level-independent or adaptive de-essers are popular because they do not need constant threshold changes.
• Live sound: engineers often use a console’s dynamic EQ or multiband compressor on the vocal channel to control harshness through the PA.

De-essers in your DAW

DAW	Stock option
Logic Pro	DeEsser 2
Pro Tools	Dyn3 De-Esser
Cubase	DeEsser
Fender Studio Pro (formerly Studio One)	De-Esser (dedicated plugin)
FL Studio	No standalone de-esser; use Maximus with its de-essing presets, or a dynamic EQ
Ableton Live	No dedicated de-esser; use Compressor with the sidechain EQ set to the sibilant range, or the high band of Multiband Dynamics
Reaper	ReaXcomp (multiband compressor) set up with a high band for sibilance
Bitwig Studio	De-Esser (stock device in the Dynamics category)

Classic de-essers and popular plugins

Product	Type	Why it matters
Orban 516EC	Hardware, 1970s	One of the first dedicated de-essers, built for recording and film.
dbx 902	Hardware, 1980	Level-independent detection; the reference design for decades, emulated today by plugins such as the Lindell Audio 902.
Drawmer MX50	Hardware	A studio staple; mixer Robert Carranza used one on Jack Johnson’s vocals (Sound On Sound).
SPL De-Esser	Hardware and plugin	Phase-cancellation approach to removing sibilance.
Weiss DS1	Digital hardware, later a Softube plugin	A mastering-grade dynamics processor with precise de-essing.
FabFilter Pro-DS	Plugin	Single Vocal and Allround detection modes, wide or split band, lookahead (FabFilter manual).
Waves DeEsser	Plugin	A long-standing, simple split-band and wideband de-esser.
oeksound soothe3	Plugin	Adaptive resonance suppressor widely used for harshness and sibilance.
sonible smart:deess	Plugin	Adaptive de-esser that targets sibilant sounds rather than a fixed band.
iZotope RX De-ess	Plugin	Restoration tool used heavily on dialogue.

How engineers use it on real records

• Tom Elmhirst on Amy Winehouse: for Winehouse’s vocals Elmhirst used the Waves DeEsser around 5.5 kHz, noting that she was not hugely sibilant and that there would probably be no more than 3 dB of de-essing (Sound On Sound, August 2007). A good reminder that effective de-essing is often subtle.
• Robert Carranza on Jack Johnson: because Johnson sang very close to the mic, Carranza needed to de-ess and used a Drawmer MX50, simply because it was in the room, even though many engineers prefer a dbx 902 (Sound On Sound, May 2008).
• Manon Grandjean on Dave’s “Funky Friday”: an SPL dual-band de-esser in the recording chain after the mic preamp to control a bright microphone, plus an SPL de-esser plugin on an effects return in the mix (Sound On Sound, June 2019).

Common mistakes and myths

Mistakes

• The lisp: too much reduction or too low a threshold removes the front edge of the s, so it sounds like th. Sibilance is part of intelligibility; the goal is to control it, not delete it.
• The wrong band: a wide band set by a preset pulls down pleasant air and breath along with the harsh peak. Find the singer’s actual problem frequency.
• Making one plugin do everything: a single heavily worked de-esser tends to pump or lisp. Clip gain plus two gentle stages usually sounds better.
• Judging in solo: a vocal that sounds dull when soloed may be perfect in the mix, and vice versa. Always check in context.
• Forgetting the effects: esses splashing in reverbs and delays are often the real problem.

Myths

• “A good de-esser can fix any recording.” Heavy sibilance often starts at the microphone: a bright condenser pointed straight at the singer’s mouth, too close. Moving the mic slightly off-axis, adding distance, or the old trick of taping a pencil vertically across the grille can prevent the problem before it exists.
• “A de-esser removes the highs from the track.” It is dynamic: it acts only for the milliseconds when sibilance crosses the threshold, and leaves the high end alone the rest of the time.
• “A normal compressor does the same job.” A broadband compressor reacts mostly to the louder vowel energy. It can even make sibilance more noticeable by reducing the vowels while the esses ride on top. A de-esser’s filtered sidechain is what makes it respond to sibilance specifically.

De-esser vs related tools

De-esser vs	The difference
Compressor	A compressor reacts to the whole frequency range. A de-esser’s detection is filtered so it reacts only to sibilant frequencies.
Limiter	A limiter caps peak level across the whole signal to prevent overloads. A de-esser targets a narrow band of vocal harshness.
Dynamic EQ	A dynamic EQ can work on any band at any frequency. A de-esser is a dedicated, simplified version tuned for sibilance, with fast timing built in.
Static EQ	A static EQ cut dulls the vocal all the time. A de-esser cuts only when sibilance happens.

Broadcast and standards

In FM radio, high frequencies are boosted before transmission by a pre-emphasis curve (75 microseconds in the Americas, 50 microseconds in Europe and most other regions) and reduced again in the receiver. Uncontrolled sibilance combined with that boost can push a transmitter towards over-modulation, which is why broadcast processors include fast high-frequency limiting, a close relative of de-essing. In streaming and TV, loudness and true-peak limits under standards such as EBU R128 and ITU-R BS.1770 are another reason to keep sharp high-frequency peaks under control before delivery.

Related terms

• Compressor: the family of processors a de-esser belongs to.
• Limiter: peak control for the whole signal.
• Transient: sibilants are short, sharp transients.
• High-pass filter: often used in a de-esser’s detection path.
• The best vocal chain: where the de-esser fits among the other vocal processors.
• Auto-Tune: another staple of modern vocal processing.

FAQ