De-essing a sibilant vocal: a practical studio workflow
Most problems in a vocal session announce themselves immediately. A hum from a power conditioner, a clipped transient, a muddy low end — you hear them in the first ten seconds and fix them before the singer settles into the song. Sibilance behaves differently. It hides inside a take that sounds honest and balanced while you stand beside the microphone, then turns abrasive the moment it meets closed-back headphones, a car stereo in traffic or a club rig. The issue is not wrongness but excess: short, bright consonants in words like song, six and she sit further forward in the mix than the rest of the voice.
The cause is physical rather than mystical. Plosives need a pop filter; sibilants need a different tool entirely, usually a de-esser — a split-band dynamic equaliser that ducks only the narrow frequency range where the s and sh sounds live, leaving the body of the voice untouched. Done well, it is nearly invisible. Done badly, it produces the lisping, hollow artefact listeners describe as smoky or underwater. At LnL Recording in Elgin, Illinois, that principle shapes the whole facility: room, microphones, outboard equipment and up to twenty simultaneous recording channels are set up so problems get solved at the source rather than patched later in post.
What follows follows the order a session actually moves in: what sibilance is, how to head it off at the microphone, how to dial in a de-esser that stays musical, how to repair a take that is already recorded, and where to stop. The examples lean towards Australian artists, because local conditions — summer air handlers, retail radio advertising rules, streaming loudness targets and the distinctly bright Australian s — shape how sibilance is heard and judged.
Where sibilance comes from
Sibilants are consonants made with a narrow constriction between tongue and teeth: s, sh, z, x and the ch sound. Unlike vowels, they are transient and broadband, with energy concentrated roughly between 4 and 10 kHz — exactly where human hearing is most sensitive to harshness and where a microphone's presence lift lives. The loudest moments in an otherwise smooth vocal can therefore sit 8 or 10 dB above the average level of the same voice.
Not every voice does this. Some singers produce a soft, breathy s that hides behind the mix; others, particularly higher voices or singers who push hard into the microphone, produce a razor-sharp one that cuts. Australian English adds a wrinkle, since many speakers use a brighter, laminal s that carries extra edge before any microphone is involved. Pair that with a condenser capsule that has a presence lift, or a bright dynamic such as the Shure SM58, and the harshness multiplies quickly.
Two misdiagnoses are common. One is hearing a harsh vocal and cutting 8 kHz globally, which dulls the voice along with every cymbal and guitar shimmer in the arrangement. The other is blaming the singer and asking them to roll the tongue back, which usually changes nothing except the tension in the performance. The reliable approach is layered: control what reaches the microphone, then control what leaves it.
Fixing it before the take
Angle beats distance. Most sibilants radiate forward from the mouth, so turning the microphone fifteen to thirty degrees off-axis, or asking the singer to angle away while the capsule still catches the chest, lowers the harshness without dulling the voice. A pop filter is essential for plosives but does nothing for s sounds, which is why the remedy is geometry rather than gauze. Height matters too: align the capsule near the sternum or just below the mouth and keep the singer off the grille's hot spot.
Microphone choice should follow the voice, not the genre. The SM58 and Beta 58A are lively and bright and reward a firm hand with a de-esser, while an RE20 or SM7B sit smoother and need far less intervention. Cardioid patterns also help, since they reject much of the off-axis energy that carries reflections and extra top end. The team at LnL Recording has written a practical guide, How to read a microphone polar pattern chart for better miking, that explains how cardioid, hypercardioid and figure-eight patterns treat off-axis sound — worth reading before you add another microphone to the locker.
When the source still resists you, capture options. Running two microphones on the same singer, a large-diaphragm condenser and a ribbon for instance, gives the mix engineer two completely different sibilance profiles to select from. Because LnL records up to twenty channels simultaneously, a full drum kit, a band with five players and a guest vocalist can all be captured in one pass, and doubling a lead vocal costs nothing but a second preamp. Buying plugins later is usually the expensive way to solve a problem that a clean take solves for free.
Room, weather and the Australian session
Air conditioning dominates a lot of Australian recording days. A compressor cycling in a Brisbane or Melbourne summer afternoon adds broadband hiss that masks fine vocal detail, and a dehumidifier running behind the vocal can colour the low mids. In Perth, recording in the late evening once the afternoon breeze drops often gives a cleaner noise floor than the middle of the day. Everyday intrusions matter just as much: the school run in a Sydney suburb at half past seven, the garbage truck on its weekly route, an inverter air conditioner on the neighbour's fence, traffic on the M1 or the Gateway Motorway. A proper noise check before the singer arrives saves hours of tidying later.
Artists travelling from Australia to record in the United States sometimes take a hotel room and set up at the bedside, which rarely survives the first listen. LnL Recording in Elgin is a purpose-built facility with professional microphones, outboard equipment, instruments, amplifiers and a custom-built digital audio workstation, and housekeeping still counts: air handler noise, street rumble, a fridge in the lounge, a thermostat switching equipment on and off. Each layer ends up on the vocal track, and de-essing cannot remove low-frequency hum; it only narrows high-frequency bands.
The Australian market shapes where the finished voice ends up too. A track may be delivered for streaming at roughly minus fourteen LUFS with a true peak ceiling near minus one dBTP, auditioned on a phone speaker with no bass at all, or broadcast on a radio network where the mix is checked against loudness guidelines. Any of those can expose sibilance that seemed perfectly civilised in the control room. Listening to the reference on headphones, a small speaker and a phone before anyone leaves is a cheap habit with a long reach.
Dialling in a de-esser that stays musical
A de-esser is a dynamic EQ restricted to a narrow band. Solo the vocal, bypass everything else, and find the exact frequency of the harshness rather than trusting a preset. Five to eight kilohertz is typical, with higher voices often sitting between six and nine. Use the spectrum view, have the singer deliver a line loaded with s sounds, and note where the peaks land and how wide they are.
Set the threshold so only the sibilant peaks are pulled down — usually two to four decibels of gain reduction on the worst consonants, with the sustained voice untouched. A ratio between three to one and six to one works well, paired with an attack of a millisecond or two and a release somewhere between fifty and 120 milliseconds. Listen to the release as much as the attack: too long and the sibilance smears into the following consonant, leaving the voice wet-sounding; too short and it chatters between syllables.
The mix control on most de-essers is a wet/dry blend, and using it well makes the processor nearly undetectable. If you can clearly hear the effect in solo, back it off until the vocal sounds like one performance again. A second pass with the dry and processed versions at matched loudness, checking for any change in vocal body, catches what repetition hides. Where takes differ, automation is the answer: ride the threshold or the gain-reduction level across the comp so a hot second chorus is controlled without flattening the whole song.
Order of operations and other tools
De-ess before you compress wherever possible. With the de-esser first, the level reaching the compressor is smoother and the compressor grabs less on consonant peaks. When compression comes first, the sibilant peaks have already been lifted, so the de-esser threshold needs to sit lower and the same settings will not transfer cleanly between takes.
Other tools help in different ways. A high-pass filter at eighty to a hundred hertz clears rumble, a gentle cut around 250 to 350 hertz reduces muddiness in a close-mic'd voice, and a band-pass sidechain can let the ducking respond only to sibilant energy. An expander with a slow release can thin the breath underneath a sibilant passage, and a brickwall limiter on the bus catches the peaks that escape. If you want a second opinion on how these dynamics tools shape a mix, an outside read on expansion and dynamic range is worth the detour.
Commercial work adds its own constraints. The Australian Consumer Law requires advertising claims to be accurate and not misleading, and retail radio codes expect limitations and conditions to be clearly audible — a mixing problem as much as a legal one, since a disclaimer buried beneath a bright mix invites trouble. Audiobooks, narration and podcasts carry their own specifications: consistent noise floor, headroom above the narration, music never louder than the spoken word, and enough headroom left for the mastering stage to work.
Repairing a take already recorded
Once the vocal is on disk, de-essing becomes the first step rather than the only one. Most digital audio workstations include a spectral editor, and it solves sibilance cleanly. Zoom into the harsh consonant, select the region, reduce the five-to-nine-kilohertz band by a couple of decibels, and shape the edges with smooth fades so the change blends into the sustained vowel on either side. Sample-accurate editing removes the edge without dulling the entire line.
Automation is the alternative. Draw a volume lane over a duplicated, compressed vocal and ride the dips by hand under the worst consonants, then automate a band-pass filter to follow the shape of each s. This takes longer than a de-esser but gives precise control, and it is often the better answer for a single hero line that needs to remain completely natural in every other respect.
Sometimes the honest answer is to comp. With twenty channels available, a session can capture ten takes of the same line and the engineer can assemble the best consonants from several of them, choosing the phrases with the softest s sounds. Arrangement also plays its part: a dense chorus with layered guitars, hats and cymbals occupying the same band as the sibilance can be tamed by moving the vocal, adding absorption or rebalancing, which often beats any amount of processing.
A checklist before the mix is signed off
- Solo the vocal and set the de-esser band to the measured frequency of the harshness, commonly five to eight kilohertz, rather than trusting a preset.
- Set the threshold so the worst consonants lose two to four decibels while the sustained voice loses nothing at all.
- Keep the attack under three milliseconds and the release between fifty and 120 milliseconds, listening for chatter and for a smeared tail after each s.
- Compare the dry and processed versions at matched loudness on headphones, small speakers and a phone before accepting the setting.
- De-ess ahead of compression so the compressor is reacting to the voice rather than to consonant peaks.
- Keep a second microphone angle or a clean comp available, because the best fix is often the take with the softer consonants.
- Leave headroom for mastering, keeping the vocal bus peaking well below zero dBFS even after the de-esser has done its work.
Those steps keep the correction honest: the singer should hear their own voice, not the processor. Everything upstream feeds the same destination, and at LnL Recording the mix, mastering and release support sit under one roof, so a corrected vocal can travel all the way to delivery with proper metadata, an ISRC and the level discipline streaming services expect. When the voice still sounds comfortable on a phone speaker in a Parramatta car park at six on a Friday evening, it will hold up almost anywhere else.
Book Your Session
Ready to record? Reach out to discuss your project, check availability, and get answers about the studio setup.
Studio Services
Professional multi-track recording with world-class gear at affordable hourly rates.
Built for Serious Sound
A studio designed around a custom PC platform — not an off-the-shelf solution — with decades of hands-on engineering experience behind every session.
Studio Gear
A comprehensive collection of microphones, preamps, outboard processing, instruments, and monitoring — all detailed on the Studio Gear page.
- AKG C414B-TLII — Large-diaphragm condenser
- Shure KSM32/SL — Studio condenser
- Shure SM57 & SM48 — Dynamic workhorses
- Shure SM81, AKG C1000S — Small-diaphragm condensers
- CAD Equitek E-100 — Supercardioid condenser
- Oktava MC012 — Multi-capsule condenser
- Sennheiser e906, E602 — Guitar cab & kick drum
- AKG D112, EV N/D 468, Audix Fusion 6
- Studio Projects C1
- Mackie MS1642-VLZ4 mixer
- Black Lion Audio Auteur Quad preamp
- Black Lion Audio B173 preamp
- Joe Meek VC1QCS channel strip
- ART TubeMP tube preamp
- Radial J48 active DI
- Rocktron Intellifex effects processor
- BBE 462 Sonic Maximizer
- Rolls RA62HA — 6-output headphone amp
Client Roster
Artists and bands who have recorded at LnL Recording.
Ready to Record?
Located in Elgin, Illinois. Reach out to discuss your next project.
Contact Us