Project
De-Ess: AUv3 Vocal De-Esser
Clean up harsh sibilance — wideband, split-band, and spectral
What it is
De-Ess is an iOS AUv3 de-esser that tames harsh sibilance — the piercing "s", "sh", and "ts" energy in vocals — before it reaches your mix. It runs inside any AUv3 host (AUM, GarageBand, Cubasis, Logic Pro for iPad, Loopy Pro, and more) with a dark, touch-first studio interface and three de-essing engines: wideband, split-band, and spectral. Sibilance detection spans 2–16 kHz, with the band for most vocals around 5–10 kHz, and threshold, ratio, and range controls shape how much it engages, alongside look-ahead timing, a Listen solo, stereo link, dry/wet, output gain, and live gain-reduction and peak metering.
Who it’s for
Vocalists, producers, podcasters, and voice-over artists on iPhone and iPad who want to clean up harsh sibilance without dulling the vocal — inline in their favorite AUv3 host.
Features
Three de-essing modes
Wideband, split-band, and spectral engines — from a fast, broad tame to frequency-specific reduction that leaves the rest of the signal untouched.
Sibilance frequency detection
Set the detection centre from 2–16 kHz; vocal "s" and "sh" sounds usually sit around 5–10 kHz.
Threshold, ratio & range
Control how much and how often the de-esser engages — from gentle taming to firm clamping of loud sibilants.
Transparent timing
Fast attack and release follow the natural shape of sibilance, with look-ahead so reduction starts before the hiss, not after it.
Listen mode
Solo exactly what the de-esser is removing so you can hear and dial in the reduction precisely.
Stereo link and mix
Process both channels with shared detection, then set dry/wet balance and output gain.
Live metering
A gain-reduction meter shows how much level the de-esser is removing, with peak meters on the input and output.
Works in any AUv3 host
Runs as an Audio Unit extension on iPhone and iPad — AUM, GarageBand, Cubasis, Logic Pro for iPad, Loopy Pro, and other compatible hosts.
Fully accessible
VoiceOver labels every control, and knobs are adjustable.
Tech
- Swift
- SwiftUI
- C++ DSP kernel
- AUv3 extension architecture
- Real-time audio
- AU parameter automation
- iPhone & iPad
- iOS 16+
Why de-ess
Sibilance is the hiss-like burst on "s", "sh", and similar consonants — the part of a vocal that cuts through a mix and fatigues the ear. A de-esser detects that band and reduces its level only when it appears, leaving the rest of the performance untouched. De-Ess makes the reduction surgical: you hear less hiss, not less voice.
Three engines
Wideband mode applies fast, broad reduction for the most transparent first pass. Split-band mode separates the sibilant band and compresses only it, preserving everything else. Spectral mode is the most surgical — it reduces the specific spectral region where the sibilance lives, useful when the vocal and the hiss share the same band.
Dial it in
Set the detection frequency from 2–16 kHz — vocal "s" and "sh" sounds usually sit around 5–10 kHz — then shape the response with threshold, ratio, and range. Fast attack and release follow the natural envelope of sibilance, and look-ahead lets the gain reduction start before the hiss. Listen mode solos exactly what the de-esser removes so you can hear what you're cutting, and stereo link keeps the image intact across both channels.
Metering and layout
A live gain-reduction meter shows how much level the de-esser is taking out, with peak meters on the input and output. A compact channel-strip view keeps the essentials reachable in an AUM slot, while iPad and wide hosts get the full studio layout. Parameters can be automated in hosts that support it.
Works where you work
De-Ess is an Audio Unit extension, so it shows up like any other effect in AUM, GarageBand, Cubasis, Logic Pro for iPad, Loopy Pro, and other AUv3-compatible hosts. Every control carries a VoiceOver label, and knobs are adjustable.