Kesko Labs DSPKESKOABSDSP
Adaptive Multiband De-Esser

AHFSC

Adaptive High Frequency Spectral Control

An adaptive multiband de-esser and resonance controller. AHFSC learns the source, proposes up to four reduction bands — sibilance, harshness, nasality, boxiness — and cuts each one in the spectral domain, only when and where the problem appears. Drag bands directly on the display; every band has its own Amount, Harmonics and self-ducking Air.

One-time purchase, perpetual license
14-day full version available
Online activation, serial-based
No iLok required
14-day money-back guarantee

Or get all six plugins in the Complete Bundle — €179 (save €136)

Two Ways In

Automatic when you want it. Multiband when you need it.

Drop it in and AHFSC works like a smart one-knob de-esser — the engine learns the source and places the reduction itself. The moment a track needs more, press ANALYZE or double-click the display and it becomes a full multiband problem-solver. Same learned model, same window, no mode switch to hunt for.

Automatic

Zero-setup de-essing

  • The engine learns the source and places the reduction region itself (AUTO)
  • One Amount slider; Tone Shift and Precision shape the character
  • Reduction happens in the spectral domain, only when sibilance appears
Multiband

Four problems, one plugin

  • Up to 4 free-range bands — not fixed crossovers: place, resize and overlap them anywhere, drawn directly on the display
  • ANALYZE proposes the layout; resonance mode cuts what classic de-essers protect as “tonal”
  • Per-band Amount, Harmonics and self-ducking Air — cut, densify and brighten independently

One visual language across both

White line = de-essing: it glows brighter the harder Amount works
Ice = Air, gold = Harmonics — every curve's intensity follows its knob
Delete all bands and the engine takes placement back automatically
Audio

Audio Examples

Comparative demonstrations of processing applied to representative source material.

Features

Technical Capabilities

Learns the voice: 3–5 second analysis builds a per-singer sibilance model

ARA integration — instant offline analysis of the whole region in ARA hosts

Adaptive tracking follows the sibilance focus in real time

Frequency-selective spectral reduction — the rest of the signal is untouched

Two engines: lookahead spectral for mixing, zero-latency for tracking and live

Enhance section rebuilds polished top end after de-essing (Air / Harmonics / Mix)

Delta monitoring — hear exactly what is being removed

Full automation support for all parameters

Signal

Learn, Divide, Conquer

AHFSC is built around a learned model of the specific voice on the track. One press of ANALYZE calibrates it in a few seconds — and in ARA hosts the entire region is analyzed offline, instantly. Reduction then happens in the spectral domain, only where sibilance actually lives, and a three-band Enhance section adds back polished top end once the harshness is gone.

Presence

Definition and closeness. Brings a de-essed vocal forward again without reintroducing edge (up to +6.5 dB).

Sheen

Polish and brightness in the classic top-end region — the shine that heavy de-essing usually costs you (up to +9 dB).

Silk

A true 40 kHz Air band, analog-matched so the rise never cramps at the top of the spectrum — the Maag/AirEQ character. At 88.2 kHz+ sessions it genuinely lifts beyond 20 kHz (up to +14 dB).

The Processing Chain

Analyze

One learning verb. A 3–5 second calibration builds a 64-point model of the voice's sibilance signature across 2–16 kHz; in ARA hosts the region is analyzed offline the moment you press it. A live-capture fallback keeps it working anywhere.

Adaptive Tracking

The detector follows the sibilance focus as it moves — different vowels, different mic distances, different phrases. The display shows the current focus frequency and gain reduction in real time.

Spectral Reduction

Gain reduction is applied per frequency inside each band, only when its problem appears. Every band has its own Amount; bands below 2.5 kHz run in resonance mode, cutting the sustained nasality classic de-essers protect as “tonal”. Tone Shift and Precision shape the whole engine's character.

Two Engines

A lookahead spectral engine for mixing quality, and a zero-latency engine for tracking and live use — switch per session, same learned model drives both.

Enhance

After reduction, the Air / Harmonics / Mix section rebuilds the top end with the three bands above — so the vocal ends up smooth and bright, not smooth and dull.

Monitoring

Delta plays only what is being removed — the fastest way to tune Amount honestly. Bypass gives an instant A/B.

Specifications

Technical Specifications

Reduction BandsUp to 4, freely placed 500 Hz – 16 kHz
Per BandAmount · Harmonics · self-ducking Air
EnginesLookahead spectral + zero-latency
Processing64-bit double precision
Latency0 samples (zero-latency engine)
Spectral Engine2048-pt STFT, 75% overlap
Sample Rates44.1 – 192 kHz (follows the host)
FormatsVST3, AU
PlatformsmacOS 11+ (Intel & Apple Silicon)
CPUIntel / Apple Silicon (native)
ActivationSerial-based, online activation