How to De-Ess Vocals Without Making Them Dull
Vocal mixing · 4 min read
De-essing works when you aim at the right frequency and take out less than you think. Find the sibilance with a narrow +12 dB boost swept between 4 and 12 kHz, set a split-band de-esser there, and pull 3–6 dB off the worst S only. One de-esser doing 10 dB always sounds like a lisp.
Sibilance lives at
5–10 kHz
Pull 3–6 dB · split it into two stages
The 60-second answer
Boost a bell +12 dB at Q 8, sweep 4–12 kHz until the esses scream, note the frequency and delete the boost. Insert a split-band de-esser there, threshold set for 3–6 dB of reduction on the loudest S and nothing on the vowels. Need more? Add a second de-esser later in the chain.
Why one de-esser doing 10 dB sounds like a lisp
An S is not a defect. It is a burst of broadband noise that tells the ear a word ended, and it carries a lot of the intelligibility you get on a phone speaker. Pull 10 dB out and the word turns into an F. The reason it needed 10 dB is usually aim: a de-esser sitting at 8 kHz on a voice whose S peaks at 6 kHz travels a long way before it touches the problem, and everything it moves on the way is air you wanted.
Broadband or split-band, and why it matters
A broadband de-esser turns the whole vocal down when it detects an S, so the vowel under that word ducks too and you hear a small hole. A split-band de-esser only turns down the band you aimed at, which is why it is the default on modern pop and rap vocals. Broadband is worth trying on a vocal already flattened by compression, where a 2 dB dip reads as a pull-back rather than a filter.
De-essing, step by step
| # | Do this |
|---|---|
| 01 |
Confirm it is sibilance S, Z, T, Ch and Sh sit above 5 kHz. Pain on vowels and belted notes is a 2–5 kHz resonance, and no de-esser touches that. Sibilance: 5–10 kHz |
| 02 |
Sweep to find the exact frequency Narrow bell, big boost, dragged slowly. Stop where the S becomes unbearable, note the number, remove the boost. Bell +12 dB, Q 8, 4–12 kHz |
| 03 |
Set the de-esser there, split band Audition the detected band alone. It should sound like esses and nothing else. If you hear vowels, narrow it or move up. Split band, listen mode on |
| 04 |
Threshold for 3–6 dB on the worst S The meter should hit on esses and stay still everywhere else. Movement on every syllable means the threshold is too low. Reduction 3–6 dB max |
| 05 |
Place it before the compressor An untamed S makes the compressor duck the whole word. De-ess first and the compressor reacts to the voice, not the hiss. tuner → de-esser → comp |
| 06 |
Automate the two lines that still fight Automate the threshold down for those instead of setting one threshold hard enough to survive the worst word in the song. Automate threshold, not output |
The trick: two small stages, one before and one after your saturation
Saturation, tape emulation, exciters and air shelves are all harmonic generators, and harmonic generators love sibilance. Feed a 6 kHz S into a saturator and it comes out with fresh energy at 12 kHz and 18 kHz that was never in the recording. So the vocal you de-essed perfectly at the top of the chain has new sibilance by the time it reaches the fader. Split the job: 2–3 dB before the compressor and saturation, then a second de-esser after them doing another 2 dB, aimed a little higher at 8–10 kHz where the new content sits. Two 2 dB moves in two places are inaudible. One 5 dB move at the end has to fight both the original S and the harmonics you generated, and that is the setting that makes a vocal sound like it is singing through a blanket.
Stop guessing the chain.
The Vocal Chain Bible is 88 real vocal chains from Grammy-winning engineers — every plugin, every setting, in order, with free alternatives for each one. 5+ years, 100+ interviews, one PDF.
Get the Vocal Chain Bible — $48 →Mistakes that kill it
- Aiming by preset. A default 7.5 kHz de-esser on a deep male voice removes air and leaves the S where it was. Sweep first, every time.
- Judging in solo. Sibilance is far more offensive soloed than it is next to hats and a snare top. Set it in context or you over-de-ess.
- De-essing after the air shelf. The shelf lifted the S by the same amount it lifted the air, so now you need a bigger cut than the problem deserved.
- Fixing a mic problem with a plugin. A bright condenser at 10 cm on a sibilant singer needs 15 degrees of off-axis angle, not 8 dB of reduction.
Keep going
FAQ
What frequency is sibilance in vocals?
Usually 5–10 kHz. Deeper voices sit at 5–7 kHz, brighter voices at 7–9 kHz and above. Sweep to find yours instead of guessing.
Should the de-esser go before or after the compressor?
Before, as a default, so the compressor is not ducking whole words because of one loud S. A second small de-esser after the saturation catches what got created there.
How much de-essing is too much?
More than about 6 dB in one stage starts to sound like a lisp. Split it across two stages, or fix the mic angle and record it again.
Who's behind this
Ivan Meshcheriakov · IvanFromRnD
5+ years ago I moved to LA with zero connections and started building from nothing — 14-hour days, meeting people and interning at studios. It took me 5+ years to meet and interview 100+ Grammy-winning engineers — then I turned it all into the Vocal Chain Bible.
Read my full story →