Which frequencies does the tool work on?

The band where sibilants concentrate, around five to nine kilohertz. Rather than filtering that range out, the tool only scales it down while it dominates the signal, so normal speech and voicing are passed through as they were. The report quotes the band along with the reduction applied.

How is this different from just cutting the treble?

A treble cut applies all the time, so a voice loses its air, presence and intelligibility everywhere. This is a dynamic processor: it watches the band's share of the total energy and ducks it only during the moments that are actually harsh. Most of a sentence is untouched, which is why de-essing sounds natural where a shelf does not.

Why do my cymbals sound slightly different afterwards?

Because cymbals put a lot of energy in the same band the detector is watching. On a full mix they can occasionally dominate it, and when they do the tool ducks them just as it would a sibilant. On speech-led material this is rarely noticeable; on a drum bus it is the reason to keep the maximum reduction low.

Can it make a lisp worse?

It can only reduce the band, never boost it, so it will not exaggerate a lisp. What it can do is make a bright voice sound dull if you push the maximum reduction too far, because some of the character of the consonants is genuinely being removed. Use the smallest reduction that fixes the harshness.

Does the band split leave any artifacts?

No. The split is done with zero-phase filtering, applying the same corners forwards and backwards, so the two halves recombine exactly with no phase rotation left over. Verified behaviour is that a high-frequency burst is attenuated while the voiced part of the signal passes byte-identical.

Sibilance Reducer

You Might Also Need

Plosive Remover

Reduce explosive breath sounds

Hiss Remover

Eliminate tape hiss and high-frequency noise

Equalizer / 3-band EQ

Adjust bass, mid and treble with a 3-band equalizer

Voice Recorder

Record your voice in browser

Sibilance Reducer softens harsh s and sh sounds by ducking only the band where they live and only while they dominate. The voiced part of the signal stays untouched, and the tool never amplifies anything.

What is Sibilance Reducer?

Load a voice recording and the tool behaves as a de-esser: it splits the signal into a frequency band and the rest, measures how much energy that band is carrying and pulls it down when it takes over the signal. Because the band is only ducked while it is dominant, ordinary speech comes through unprocessed and cymbals in a full mix mostly survive.

The controls and the readout:

A sensitivity slider from 10 to 30 dB, defaulting to 20, expressed as the share of energy the band must carry before a cut is applied.
A maximum reduction slider from 2 to 15 dB, defaulting to 6.
A report line giving the number of sibilance events softened and the maximum reduction applied.
The affected band, which the tool works in around the 5 to 9 kHz region where sibilants concentrate.
A WAV download named after the source with a deessed suffix.

The band split is done with zero-phase filtering, which means the same corner pairs are applied forwards and backwards so no phase shift is left behind by the split itself. That matters because the two halves of the signal are recombined afterwards: a conventional split would leave the high band rotated in time relative to the low one and the recombination would no longer be transparent. The detector compares the band's energy against the total, and when the share crosses your sensitivity threshold the band is scaled down by up to the maximum reduction, smoothly and monotonically, so there is no filter ringing added at the moment of the cut. Verified behaviour is that a six kilohertz burst is attenuated by at least twenty percent while the voiced part of the same signal is byte-identical, that normal speech is untouched, and that the processor never boosts anything. That last property matters in practice, because a de-esser that overshoots its target leaves the consonant sounding worse rather than merely softer.

What it does not do:

Cymbals and airy percussion share the band, so on a full mix the duck can still catch them when they dominate on their own.
A lisped or unusually bright voice may need a low maximum reduction to avoid sounding dulled rather than fixed.
It cannot undo distortion that a sibilant already caused in the microphone chain.
The band is fixed around the sibilant region, so it is not a general purpose dynamic equaliser.
Detection works per channel, so an aggressive sibilant that appears in only one channel of a stereo pair is handled only there.
It processes offline with no visual map of where events fired, only a count in the report.

Use it on close-miked spoken word, where bright consonants are the most common complaint, and on any vocal take where the esses cut through the mix in an unpleasant way. Start at the defaults, listen to a sentence with several s sounds, and lower the maximum reduction if the voice starts to lose its presence rather than its harshness, since the aim is to remove the sting and not the character.

How to use Sibilance Reducer

Load a recording with prominent s and sh sounds and leave the sensitivity at its default.
Set the maximum reduction, beginning at 6 dB and going lower if the voice starts to sound dull.
Run the pass and read the report for the number of sibilance events and the maximum reduction applied.
Check a sentence with several sibilants against the original, then download the deessed WAV.

When to use Sibilance Reducer vs related tools

Sibilance Reducer targets the bright, harsh consonants of close-miked speech, which is a narrow job by design. Plosive Remover addresses the opposite end of the same problem, taming low-frequency blasts instead of high ones, and Hiss Remover takes down broadband top-end noise rather than concentrating on a band. When the whole recording needs its balance changed rather than one problem band ducked, Equalizer gives you the bass, mid and treble shelves, and Voice Recorder is where a take with better microphone placement comes from.

Privacy & Security

This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.