Why use a de-esser instead of EQing out the esses?

A part-time problem

Sibilance is a burst of energy, usually between 5 and 9 kHz, lasting around a tenth of a second. Between those bursts, the same frequency range carries the clarity of the voice. A fixed cut deep enough to tame the esses leaves everything else dull.

How a de-esser works

It is a compressor that listens to one band. A detector measures the level between two frequencies. When that level passes a threshold, the device turns the band down, and when the consonant ends it lets go. Nothing happens on the words between.

Settings

• Frequency. Sweep until the esses are loudest in the detector. Lower voices sit nearer 5 kHz, higher voices nearer 8 kHz.
• Threshold. Low enough that every hard 's' triggers it, high enough that vowels do not.
• Amount. 3 to 6 dB of reduction on the worst ones. Much more and the singer starts to lisp.

Compression and bright EQ both make sibilance worse, so the de-esser usually goes after them in the chain.

When EQ is the right tool

When the harshness is constant: a bright microphone, a resonance in the room, a peak that is there on every word. That is a tonal problem and a static cut fixes it. The test is whether the problem comes and goes with the consonants.

In Zammoth

The De-esser's band can be set anywhere from 2 to 12 kHz and starts at 6 kHz. Its display is a six-second tape of the detector band with every catch marked, so you can see which esses it took and by how much. Hold SC to hear the detector band alone, and switch Wide on to turn the whole signal down instead of the band. The engineer measures sibilant bursts phrase by phrase and can de-ess only the stretches that need it. See the De-esser reference.