De-esser: tame harsh S sounds
S sounds that hiss or spit? fix.fm measures them against your voice and turns down only those moments. Free online de-esser, nothing uploaded.
UPDATED
Never uploaded. Everything runs in your browser.
In short
Harsh S sounds, or sibilance, are S and SH sounds that hiss or spit because the top of the voice is too loud, from a bright mic, a close mouth or a treble boost. fix.fm measures the S sounds against your own vowels and turns the top end down only in the moments an S stands out. On our bright-mic sample the loudest S sounds went from -6.3 to -13.6 dB against the voice.
What's happening
Sibilance is when the S, SH and soft C sounds come out too loud: they hiss, whistle or spit, and on headphones they can feel like a needle in your ear. It usually comes from a bright mic, talking too close or straight into it, or a treble boost added while editing.
fix.fm measures the loudest S sounds against the loudest vowels in the same voice, so a naturally bright voice isn't punished for being bright. Its warnings start when the S sounds come within 11 dB of the vowels; it calls them severe when they come within 3 dB. The fix is a de-esser: it turns down the top of the spectrum only during the S sounds, lets go quickly, and leaves every other moment exactly as it was.
How far the de-esser goes
On the bright mic, the loudest S sounds fell from -6.3 to -13.6 dB against the vowels while the words under 3 kHz moved 0 dB. A harsh boost went from -0.3 to -12.8 dB, and a darker voice with the same boost was left alone. Measured on with the samples described in how we test.
The de-esser judges S sounds against the voice they belong to, not against a fixed level. The three cases below are clean readings with a treble boost added in software, so we know what was there and can measure what the de-esser takes away.
Two clean readings with a treble boost added in software, then run through this page's de-esser at full amount. Every clip is matched to -23 LUFS so only the S sounds differ.
Clean reading
The reference. Voice A as recorded, with smooth S sounds.
Bright mic, +8 dB
Like a bright condenser up close: every S and SH comes out sharp and a little hissy.
Bright mic, after the de-esser
The S sounds sit back down. The rest of the voice keeps its brightness, since only the S moments are turned down.
Harsh boost, +14 dB
A heavy treble boost: the S sounds spit and are louder than the words around them.
Harsh boost, after the de-esser
Much calmer. The de-esser cuts deeper and more often here, so listen for a slight lisp on the longest S sounds.
Darker voice, +8 dB
Voice B with the same boost as the bright mic. Brighter than before, but the S sounds still sit under the words.
Darker voice, after the de-esser
Identical to the recording. The check reads this voice's S sounds as smooth, so nothing is touched.
Public-domain LibriVox voices. The harshness is a high-shelf boost added in software, not a different mic, so the clean version really is what was there before.
| Case | The check | Above 4.5 kHz | Under 3 kHz | Samples changed |
|---|---|---|---|---|
| Clean reading | -14.8 dB, fine | 0 dB | 0 dB | 0% |
| Bright mic | -6.3 dB moderate to -13.6 dB fine | -3.7 dB | 0 dB | 13.2% |
| Harsh boost | -0.3 dB severe to -12.8 dB fine | -6.7 dB | 0 dB | 26.1% |
| Darker voice, same boost | -19.1 dB fine to -19.1 dB fine | 0 dB | 0 dB | 0% |
The bright mic is the case the de-esser is for. The check reads -6.3 dB, moderate; after the fix it reads -13.6 dB, fine, close to the clean reading's -14.8. The top of the spectrum comes down 3.7 dB over the whole file, the band where the words live moves 0 dB, and 13.2% of the samples changed. That is the S moments and their short release. Everything between them is the sample you recorded.
The harsh boost starts severe, at -0.3 dB, and ends fine at -12.8 dB. To get there the de-esser cuts deeper and more often: 26.1% of the file and 6.7 dB off the top. On a boost this heavy the longest S sounds can start to sound a little lispy. If they do, turn Amount down until they don't; the treble boost itself is better undone in your editor.
The darker voice got the same +8 dB above 5 kHz as the bright mic, but its recording has little above 7 kHz to lift, so its S sounds still sit well under its words: -19.1 dB, fine. The de-esser returns it unchanged. A fixed threshold would have cut both voices the same; this one only acts on a voice whose S sounds stand out from that voice.
What this does not prove: a treble boost is not a real bright mic. Real harshness can sit lower, near 3 to 4 kHz, where this check doesn't look, and a lisp or a whistle on S is a different sound again. If the whole voice is thin rather than the S sounds, the clarity fix is the wrong direction and an EQ is the right one. Reducing sibilance covers mic placement, and how we test has the pass marks behind the check.
How it works here
- 01
Measures the S sounds
It compares the loudest S and SH sounds with the loudest vowels in your own voice, so a bright voice and a dark one are each judged on their own terms.
- 02
Turns down only the S
When an S stands out, the top of the sound is turned down for that moment and lets go right after. Everything under 3 kHz, where the words live, is left alone.
- 03
Stays on your computer
Your file is never uploaded. The de-esser runs in this browser tab.
Questions
What causes harsh S sounds?
A bright condenser mic, talking very close to the mic or straight into it, or a treble boost added in editing. Some voices are naturally sharper, and MP3 or video compression can make S sounds spit more.
What does a de-esser do?
It listens for S sounds and turns the top of the voice down only while one is too loud. An EQ cut does the same all the time, which dulls every word to fix a few S sounds.
Will it make my voice dull or lispy?
It shouldn't. The cut only happens in the moments an S stands out, and only above about 4 kHz. On our bright-mic sample the words under 3 kHz didn't change at all. Turn the amount down if the S sounds start to lisp.
Why didn't it change my recording?
Because the check found the S sounds already smooth against your voice. fix.fm only de-esses when it measures a problem, so a clear recording comes back exactly as it went in.
Is my audio uploaded anywhere?
No. The file is read and fixed inside this browser tab and never leaves your computer.
How do I stop harsh S sounds next time?
Speak across the mic rather than straight into it, back off a few centimetres, and use a pop filter. Skip the treble boost, or add a de-esser in your editor before it.