De-Esser
Tame harsh “s” sounds. See where they sit, and hear what was removed.
Load a vocal or dialogue track, point the Focus at the sibilant region, and the De-Esser turns that region down only while a sibilant is actually happening. Compare with the original, listen to exactly what was taken out, and download the result as a WAV. Your file never leaves your device.
This browser doesn't support the audio features this tool needs (Web Audio). Try a current version of Chrome, Edge, Firefox or Safari.
DROP YOUR TRACK
or click to choose a file
WAV, MP3, AIFF or FLAC, up to 8 minutes. Everything happens in your browser: the file is not uploaded and nothing is stored. Privacy notice.
Reduction over time
How far the Focus region was turned down, moment by moment (0 at the top). Click or drag to jump to a spot.
Starting points
Controls
Download
The download is the exact audio you are hearing in Processed — the same processing, the same settings, 24-bit WAV. It is created in your browser, so there is no upload and no wait for a server.
A detector, a reducer and a way to check both
- Listen in a band. A detector filters the track to the Focus region (Focus is its centre, Range its width) and follows its level.
- Compare with this track. Reduction starts when that level rises above a threshold set from the track’s own sibilant peaks, using Sensitivity. Above it, the reduction grows with a soft knee up to the Reduction limit.
- Reduce only that region. In Split mode the reducer subtracts part of the Focus band from the signal, so the rest of the voice is untouched. Attack and Release shape how the reduction moves.
- Check it. Switch between Original, Processed and Delta. The spectrum shows the average before and after; the strip below it shows when reduction happened.
The spectrum lines are measured from your file and from the processed result. The shaded area between them is the average reduction across the whole track, so it looks smaller than what you hear on an individual sibilant.
Where to start
The four starting points only set the controls above to sensible positions. They do not analyse your voice and they are not guaranteed to suit it. Adjust Focus until the shaded region sits on the bright hump in your spectrum, then raise Sensitivity until Delta mostly contains “s” and “t” sounds and very little else.
Gentle Vocal
Light touch on a mostly clean vocal.
Bright Vocal
Wider and deeper for a bright or close-miked vocal.
Harsh Sibilance
Narrower, stronger control for piercing “s” sounds.
Dialogue
Lower focus, broad range, relaxed timing for speech.
Related tools
If the whole voice is too bright rather than just the sibilants, shape it with EQ Studio. To see what is dominating a track before you decide, use the free Frequency Analyzer. If a track has peaks that come and go across the spectrum rather than just in the sibilant region, try the Resonance Suppressor.
Questions
Is my track uploaded?
No. The file is decoded, processed, played and encoded to WAV inside your browser. Nothing is sent to our servers, so there is no upload wait and nothing to delete afterwards.
Will the download sound like what I hear?
Yes. Processed is a real, fully rendered version of your track and the download is exactly that audio as a 24-bit WAV. The Match loudness switch only changes how loud the Processed side plays while you listen; it never changes the file.
What is the difference between Split and Wide?
Split turns down only the Focus region while a sibilant is detected, so the rest of the voice stays as it was. Wide turns the whole signal down instead, which can sound more natural for some dialogue and very bright sources but also dips the body of the voice for a moment.
What does Sensitivity do?
The detector measures the level in the Focus region across your whole track. Sensitivity sets how far below that track's own sibilant peaks reduction starts, which is why the same preset behaves similarly on a quiet and a loud recording. Reduction caps how far the region may be turned down, whatever the Sensitivity.
What is Delta?
Delta plays the original minus the processed audio: exactly what the De-Esser removed. If you hear words or syllables in it, the Focus or Range is probably too wide. It is the real difference signal, at its true level, with an optional +12 dB boost to hear quiet removals.
Is this an AI tool?
No. It is conventional signal processing: a band-pass detector, a gain computer with a soft knee, attack and release smoothing, and a band-limited reducer. The presets only set the same visible controls. There is nothing hidden behind them.
What can it not do?
It cannot rebuild a distorted or clipped vocal, remove noise or repair a recording that is harsh across its whole range. If harshness is not tied to sibilant sounds, try a gentle cut in EQ Studio instead. Files are limited to 8 minutes in the browser.