Start by sweeping a narrow detector between roughly 5 and 10 kHz to find the exact "ess" peak, then set the threshold so the gain-reduction meter shows about 3–6 dB on the loudest sibilants. If the vocal starts to lisp, back off. Use a fast attack, a short release and around 10 ms of lookahead where available, and put the de-esser before any saturation.
TL;DR:
- Sweep a narrow detector from 5–8 kHz for lower register voices and 6–10 kHz for higher register voices, then confirm the peak across several phrases.
- Aim for 3–6 dB of reduction on the loudest sibilants; if levels approach 10 dB, check microphone choice, placement, or equalization before increasing processing.
- Place the sibilance processor before saturation; use it before compression when esses trigger pumping, but add a later pass if compression raises sibilance.
- Choose wideband processing for solo vocals, split band for dense mixes, and spectral or dynamic tools for moving resonances or multiple problem frequencies.
- Lower isolated S and T spikes with clip gain and short crossfades, then let a gentle processor handle the remaining sibilance.
Table of Contents
- How do you tune a de-esser in minutes?
- Where is the sibilance, and how do you find it?
- Where should the de-esser sit in the vocal chain?
- Which de-esser mode suits your vocal?
- What do the controls do, and what are safe starting values?
- When should you use clip gain, automation or both?
- How do you fix lisping, the wrong band and pumping?
- Why can you trust this guide, and how do we verify settings?
- What is the one rule that matters most when de-essing?
- Want data-driven feedback on your de-essing and vocal balance?
- FAQ
- Sources
How do you tune a de-esser in minutes?
Here is the quick version we use at the console. Every step is about listening, not just watching the meter, and our de-esser settings for vocals starting points come from a simple principle: the detector should react to the consonant and ignore the vowel.
- Audition the detection band. Solo the detector (most de-essers have a listen or audition switch) and sweep a narrow band until the "S" and "T" sounds jump out at you.
- Set the threshold. Pull it down until the meter moves mainly on sibilants and stays quiet on vowels. If it twitches on the "ah" of a held note, the band is wrong or the threshold is too low.
- Dial the range or amount. Aim for about 3–6 dB of gain reduction on the loudest esses, then listen for lost intelligibility. If the word "sit" starts to sound like "this", ease off.
- Refine the timing. Set a fast attack, a short release and try lookahead. Then automate the threshold or use clip gain on the few outliers the plugin misreads.
Our approach matches the step-by-step method from inotrak, which describes a de-esser as a compressor that listens only to the sibilant band and treats 3–6 dB on loud esses as normal.
Pro Tip: Bypass the de-esser every few seconds while you dial it in. Your ears adapt quickly to a duller vocal, and the A/B check keeps you honest.
Where is the sibilance, and how do you find it?
Sibilance does not live at one fixed frequency. Typical vocal sibilance falls between 5 kHz and 8 kHz for lower-register voices and between 6 kHz and 10 kHz for higher-register voices, and the same source says to sweep the detector to find the exact resonance rather than trusting those ranges blindly. Treat them as a map, not a destination.
Here is the sweep method we recommend:
- Start in the range for the voice. Begin around 5–8 kHz for a deeper voice, 6–10 kHz for a higher one.
- Use a narrow Q while hunting. A tight band makes the peak obvious; a wide band smears it.
- Listen for the sharpest "ess". The offending frequency is where the sibilance sounds most piercing, not just loudest.
- Check the T, CH and SH sounds. Narrow the band until those consonants stay natural. If they get damaged, you are cutting too wide or too low.
- Re-check on the whole take. Mic choice and singer distance change where the peak sits, so confirm on several phrases rather than one.
If you find two separate peaks, use two bands or move to a multiband compressor or dynamic EQ. Those tools let you focus on a narrow offending band, for example near 8 kHz, while keeping the vocal's presence intact, and auditioning the band in solo is essential to avoid stripping out useful harmonic information. Our guide to dynamic EQ goes deeper on that route.
Professionals consider 3 to 6 dB of gain reduction on the loudest sibilant peaks a normal range. If you keep landing well above that, the cause is probably upstream rather than in your plugin settings.
Some stock de-essers shortcut the hunt with Male and Female presets that shift the detector centre between roughly 6 kHz and 7 kHz, which Icon Collective notes as a useful starting point. Sibilance also tends to wander when the singer moves closer to or further from the mic, so if a recording is harsh from the start, check mic technique first. A placement reference such as this microphone placement infographic is a handy reminder of how much position matters, and our broader guide to sibilance covers recording-stage fixes.
Where should the de-esser sit in the vocal chain?
Chain order is where many engineers get tangled up, so here are the rules we follow:
- De-ess before saturation or distortion. Saturation generates new high-frequency content, so a de-esser placed after it is chasing harmonics you just created. Inotrak makes the same point about placing it ahead of saturation.
- Before the compressor for inconsistent performances. If sibilant peaks make the compressor pump or grab, de-essing first lets the compressor react more predictably.
- After the compressor when earlier processing boosts the esses. Heavy compression and make-up gain can lift sibilance, so a late de-esser catches what the chain amplified. Icon Collective notes that most experts favour a late position, with the pre-compressor spot kept for uneven singers.
- Use an external side-chain on distorted vocals. Feed the detector a clean copy of the vocal so it triggers on the real sibilance, not on the grit.
- Try mid/side where it exists. Mid-only suits a lead vocal; side-only can calm panned backing vocals without touching the centre.
Plenty of mixes use both positions: a light de-ess early to protect the compressor, and a gentle pass later for what remains. Our guide to vocal compression shows how the two processors interact.
Which de-esser mode suits your vocal?
Different modes solve different problems, and picking the most transparent one saves you from over-processing.
- Wide-band ducks the whole signal when an ess appears. It is often the most natural choice on a solo vocal, because nothing sounds filtered.
- Split-band only attenuates the high frequencies, so low-mid warmth stays untouched. That makes it a good fit for dense mixes.
- Spectral processors track resonances and remove only the offending narrow content. They are excellent for wandering or whistle-like sibilance, and Artists in DSP's comparison notes that modern resonance suppressors such as Soothe2 follow non-stationary resonances more transparently than fixed-band tools.
- Dynamic EQ and multiband compression shine when sibilance overlaps other tonal problems or several regions need treatment. A multiband compressor can act as a de-esser by narrowing the crossover of the processed band, and Sound On Sound suggests muting and unmuting to confirm coverage before you commit.
Here is the rubric in one breath: quick natural fix, go wide-band; dense mix, go split-band; stubborn or moving resonance, go spectral or dynamic.
Test both Wide and Split Band on the same phrase, because they can sound different in ways a meter will never show. Icon Collective recommends exactly that comparison.
Pro Tip: Switch modes mid-loop without stopping playback. The difference in how the "t" and "s" land is easiest to hear when you hear both back to back.
What do the controls do, and what are safe starting values?
This is the control-by-control checklist we reach for. Use it as a starting point and let your ears overrule it.
- Threshold: lower it until the meter moves mainly on sibilants. If the detector misfires on some phrases, automate the threshold phrase by phrase instead of treating the de-esser as set and forget.
- Amount or range: aim for roughly 3–6 dB on the loudest sibilants. Two gentle passes of about 2–3 dB each usually sound smoother than one deep cut.
- Attack and release: a fast attack and a short to medium-fast release keep the reduction tied to the consonant. Too slow a release and the following vowel gets dulled.
- Lookahead: start around 10 ms. FabFilter's documentation says values around 10 ms frequently catch S and T transients while preserving some of the initial transient for natural character.
- Q or bandwidth: go narrow for a whistle-like peak, then widen slightly so you do not cut intelligibility.
- Stereo link and mid/side: link channels for a centred lead; for backing vocals, side-only processing can tame sibilance without dulling the lead's centre image.
- Detector audition: keep it on while you sweep and switch it off before judging the result.
A de-esser reduction of 10 dB or more often indicates an upstream problem, such as mic choice, pop filter or EQ. If you are pinned that hard, fix the source before reaching for more processing.
Advanced tools differ in how much they help. Some high-end de-essers offer auto-threshold, male and female centre presets and clever detector designs that cut down on manual riding, but Sound On Sound's review of FabFilter Pro-DS still stresses careful audition and fine-tuning. Automation does not replace listening.
When should you use clip gain, automation or both?
The most transparent de-essing is often not a plugin at all. There are two broad approaches: manual clip gain, which is very precise but time-consuming, and automatic de-essing, which is fast and effective. iZotope's vocal editing guide notes that manual edits need crossfades to avoid clicks and can produce excellent results on critical material.
Here is the hybrid workflow we like for lead vocals:
- Find the worst offenders. Scan the waveform for the loudest "S" and "T" spikes, usually visible as dense bursts of high-frequency energy.
- Cut them with clip gain. Split the clip around each sibilant, lower it by a few dB and add short crossfades at both edges so nothing clicks.
- Insert a gentle de-esser. Let it handle the sibilance that remains, keeping reduction light.
- Automate where needed. Raise the threshold or bypass the processor on quiet passages so soft phrases do not turn dull.
How do you fix lisping, the wrong band and pumping?
When a de-essed vocal sounds off, work through these quick diagnostics:
- Lisping or "thick" consonants: the detection frequency is usually too low or the reduction too deep. The first fix is to raise the detection frequency before reducing the amount. Accept a touch of sibilance rather than losing intelligibility. Our vocal intelligibility guide shows how to restore consonant clarity.
- Wrong band: switch the audition mode on and confirm the detector fires on the consonant, not on harmonic vowel energy. If it triggers on a vowel, move the band or narrow the Q.
- Pumping from the compressor: de-ess before the compressor, or manually reduce the offending spikes with clip gain so the compressor never sees them.
- Reduction stuck above about 10 dB: revisit mic choice, singer distance and any pre-EQ that is pushing the highs. The problem is almost certainly at the source.
Why can you trust this guide, and how do we verify settings?
Every number in this article comes from the sources linked in the text, and the workflow reflects how we approach de-essing in practice. We built Aubiomix as an online mix analysis app: you upload your song and get detailed feedback with actionable steps across more than 15 areas of mixing and mastering, including compression, EQ and frequency masking.
For de-essing, a review gives you a second opinion. You can see exactly what each report covers in our evaluation framework.
What is the one rule that matters most when de-essing?
Intelligibility comes first. Every de-essing decision should leave the words easier to understand, not just the highs softer. I would rather leave a little sibilance in than hear a singer start to lisp.
My practical rule: favour small, reversible moves. Switch on the detector audition, find the peak by ear, then make two light passes instead of one heavy cut. If you have to wonder whether it is too much, it is.
— AubioMix
Want data-driven feedback on your de-essing and vocal balance?
If you would like a measured second opinion on your vocal, we offer mix reviews built for exactly this. Here is what you get relative to this article:
- Detailed feedback on your mix.
- Actionable feedback to help you assess your vocal balance.
- Stepwise actions you can apply in your session straight away.
A single mix review costs £4.99 one-off, and the Professional plan is £19.99 per month; both are listed on our pricing page. Upload your mix, read the report and take the next step with your vocal today.
FAQ
What is the best EQ setting for vocals?
There is no single best setting, because the right EQ depends on the voice, the mic and the mix. Fix problems first with small cuts, then add gentle presence, and handle sibilance with a de-esser rather than a broad high cut. A sweep of the detector will tell you where the harsh energy actually sits.
How do you get rid of sibilance in vocals?
Find the offending band by sweeping a narrow detector between about 5 and 10 kHz, then set the threshold so the meter moves on the esses. Aim for 3 to 6 dB of gain reduction and use clip gain on the worst spikes. If you need much more reduction than that, check the mic and its placement.
What is the best de-esser plugin for vocals?
It depends on the job. Wide-band tools often sound most natural on solo vocals, split-band tools suit dense mixes, and spectral processors handle wandering resonances most transparently, as Artists in DSP's comparison describes. Choose the mode first, then the plugin that does it well.
Should you use a de-esser before or after AutoTune?
Pitch correction can change the character of high frequencies, so many engineers de-ess lightly after it to catch any sibilance it emphasises. If the raw take is very sibilant, a gentle de-ess before can also help the tuning detect pitch more cleanly. Audition both and keep reduction light at each stage.