Air-G Dess
User Manual
Version 1.0
De-esser based on the Airwindows DeEss algorithm.
Original source code released under an open licence (MIT) by Chris Johnson.
User Manual
Version 1.0
De-esser based on the Airwindows DeEss algorithm.
Original source code released under an open licence (MIT) by Chris Johnson.
Air-G Dess is a de-esser: a tool that turns down the sibilant sounds in a voice — the «esses», «shes» and «tees» («sss», «shh», «ts») that tend to come out too loud and harsh, especially with bright or condenser microphones. The de-esser tames them without touching the rest of the voice.
What makes Air-G Dess special is how it detects sibilance. It doesn't simply listen to «a band of highs» like a typical de-esser: it analyses the shape of the waveform. A real ess is a sound that reverses itself many times per second, fast and almost chaotically; Dess looks for exactly that pattern. That's why it is so precise: if something is bright but isn't really an ess (a cymbal, the air of the voice, a harmonic), it leaves it alone.
Under the hood it uses the Airwindows DeEss algorithm, created by Chris Johnson and released under an open licence.
Raise INTENSITY until you hear it working on the esses; use MAX DEESS to decide how far the ess is pulled down (without erasing it completely); and FREQ to choose the «colour» of the sibilance. Watch the DEES meter: it should move only on the esses and stay still the rest of the time.
Air-G Dess interface.
The plugin is resizable (keeping its aspect ratio). Double-clicking a knob returns it to its default value.
The big knob is the main control. At 0 the de-esser does nothing. As you raise it, Dess becomes more sensitive: it spots sibilance more easily and ducks it harder. This is the knob you use to «find the spot»: raise it until you hear the esses calm down. The exact value is shown below it.
It sets a limit on how far the ess can be pulled down, and works like a «how much effect» control (similar to a dry/wet). It is shown in dB, from −48 to 0:
It defaults to −24 dB, a middle ground: usually you do not want to remove the ess entirely (it would sound as if the person can't pronounce the «s»), only tame it. Think of it this way: INTENSITY decides when it acts and MAX DEESS decides how far it goes.
It defines which part of the sibilance it works on, on a LOW to HIGH scale:
Beyond taming the ess, FREQ is a way to add character: bringing it toward LOW lets the lower part of the ess through and removes only the bright edge, so the voice starts to sound as if recorded with a dynamic microphone (warmer) instead of a sharp condenser.
FREQ is shown in zones (LOW / LOW-MID / MID / HIGH-MID / HIGH) and not in Hz: the internal filter shifts on its own depending on the signal, so a fixed value in Hz would be misleading.
INTENSITYHow hard the de-esser hunts for esses. At 0 it does nothing; higher = it acts more easily and harder. · MAX DEESSLimit on ess reduction (−48 to 0 dB). 0 = no effect; more negative = more reduction. Defaults to −24 dB. · FREQThe colour of the ess, from LOW to HIGH. Toward LOW the voice sounds warmer (dynamic-mic style).
The black panel on the right has two bars that share a real numeric dBFS scale (0 at the top, down to −48 at the bottom). They are faithful: they read the signal directly, not an approximation.
The left bar shows the plugin's output level, filling from the bottom up (green → amber → red). A white line marks the recent peak (peak-hold), so you can see how close to 0 dBFS you are getting.
The right bar shows how much the ess is being pulled down right now, in dB, drawing from the top down from 0. It rises only when Dess is catching sibilance. It is your visual guide: if the DEES bar flicks up right on the esses and stays still the rest of the time, you have it well dialled in.
Ideally DEES moves only during the esses. If it moves all the time, you are over-ducking (lower INTENSITY or raise FREQ); if it barely moves when there is sibilance, there isn't enough effect (raise INTENSITY or lower MAX DEESS). Use OUT so you don't exceed 0 dBFS at the output.
The classic use: tame the esses of a lead vocal without it being noticeable. Find the spot with INTENSITY, then ease off with MAX DEESS and FREQ until it sounds as if the microphone had been perfect from the start — no harsh esses, but without losing the highs of the voice.
With sharp microphones the esses cut. Raise INTENSITY and aim FREQ toward HIGH to take the extreme edge off the sibilance.
Bring FREQ toward LOW to let the lower part of the ess through and remove only the brightness: the voice sounds warmer, like a dynamic microphone. Beyond correction, it is a tone tool.
In spoken voice, sibilance tires the listener. Gentle de-essing (moderate INTENSITY, MAX DEESS not overdone) makes a reading more comfortable over long stretches.
1) Exaggerate for a moment (FREQ toward LOW, MAX DEESS toward −48) and raise INTENSITY until you clearly hear where it hits. 2) Ease off: raise FREQ and lower MAX DEESS until the effect is invisible. 3) Confirm with the DEES bar: it should move on the esses and stay still on the rest.
Air-G Dess uses the DeEss de-essing algorithm by Airwindows, created by Chris Johnson, under an interface with an intensity knob, a reduction limit (MAX DEESS), an ess colour control (FREQ) and two digital dBFS meters.
The original algorithm is the work of Chris Johnson / Airwindows and is used under the terms of his open licence (MIT). The original source code is publicly available.
To Chris Johnson, for sharing his de-essing algorithm under an open licence, and for the documentation that served as the basis for this manual.
Air-G Dess v1.0 — Air-G Audio. The de-essing algorithm is the work of Chris Johnson / Airwindows and is used under the terms of his open licence (MIT).