Sound
Sound by Human Engine Labs

It adapts to every take. You make the call.

Adaptive audio plugins that read the signal in real time and tune themselves to it, so one intuitive control does what used to take an afternoon of settings. Human engineering, supported by adaptive technology.

Signal / oscilloscopeLive
100 Hz1 k4 k8 k16 kHz
Perpetual licences
Measured on real vocals
VST3 · AU · CLAP · Standalone
01

The approach

Deep tech, analog soul.

Under the hood, adaptive AI: deep learning, neural models, and real-time sound analysis. On the surface, analog warmth and character. And in every detail, the care of people who take real pride in the work.

A

AI that listens, in real time

Deep-learning classifiers and live sound analysis read what is actually in the signal, the phoneme, the voice, the room, and adapt to it, so one intuitive control is all you touch.

B

Analog warmth, on purpose

Modelled saturation and tube and tape textures give the result character, not just correction, and leave room to be creative rather than clinical.

C

Made with obsessive care

Every meter shows something real, and every number is measured on real material. We build these the way we would want them built for us.

03

The engineering

Three lanes. No lane decides alone.

The de-ess decision runs through three dissimilar lanes, the command-and-monitor pattern from flight-control computers. No single lane fault can cause a wrong cut.

COMMAND
Deterministic DSP detector

Measures sibilant energy per block. Fast and deterministic, but it can’t tell an “s” from a plosive, so it never decides what a sound is.

MISSION
AI sequence classifier

Classifies the sound. It can veto the cut to protect a plosive or fricative, but it can never trigger one.

CONTROL
Temporal monitor

Requires the veto to hold across consecutive windows, so one wrong frame can’t trip protection or flash a false letter on the meter.

signalCOMMANDdeterministic DSP · drives the cutMISSIONAI classifier · can only vetoCONTROLtemporal monitor · veto must persistCommitonly where lanes agree
Three dissimilar lanes, the command-and-monitor principle from flight-control computers. The AI never drives the cut; it can only protect.
04

Measured, never faked

Measured on real vocals.

Validated end to end on real vocal recordings, not synthetic test signals, which would let us print kinder numbers than these.

~5 dB
Reduction on harsh esses
±0.04 dB
Rest of the band vs. untouched
2.5 %
One CPU core · RTF 0.025
22
Validation gates, all green
05

Pricing

Perpetual licences.

A one-time purchase, never a subscription. The free tier is licensed for commercial work.

Sibilance Open
For the curious.
Free
Unlimited machines
  • +Full single-authority de-esser
  • +Δ Listen · Auto Gain
  • +Licensed for commercial work
Sibilance Pro
For the committed.
$79 once
Perpetual · 3 machines
  • +The full macro set + analog finish
  • +Mid/Side, De-Lisp, sidechain
  • +All presets, A/B
Thunder Clarity
For the expert.
$149 once
Coming soon · 3 machines
  • +Three-fader neural separation
  • +Classic + neural cores
  • +Delivery metering
The Clarity Series
For the steward.
$189 once
Coming soon · both instruments
  • +Sibilance Pro + Thunder Clarity
  • +Every future point release

The list

Get the first release.

Sibilance ships first; Thunder Clarity follows. The list gets the release and the intro price, and nothing else.

Release news only, unsubscribe any time. Privacy.

Release notes only.

FormatPlatform
VST3macOS · Windows
Audio Unit (AU)macOS
CLAPmacOS · Windows
StandalonemacOS · Windows

Apple Silicon & Intel · JUCE 8 · C++20