De-Esser

A de-esser is a frequency-specific compressor. It only listens to a narrow band — typically 5–10 kHz — and only ducks the signal when energy in that band exceeds a threshold.

That makes it the right tool for taming the harsh "S" sounds (sibilance) that almost every vocal recording suffers from at some point.

How It Works

A de-esser splits your signal into two paths. One path is a side-chain detector that listens to a specific frequency range — the band where sibilance lives. The other path is the audio you actually hear.

When sibilance occurs, the detector triggers a compressor that reduces the gain on the audio path. As soon as the sibilance passes, the gain returns. The result: the voice sounds the same except in moments where it would have been harsh.

Where Sibilance Lives

Different voices have sibilance in different places:

• Male voices — often 5–7 kHz.
• Female voices — often 7–9 kHz.
• Younger voices — sometimes higher, 8–10 kHz.
• Recordings on bright condensers — anywhere from 4 kHz to 12 kHz.

The right starting move is to solo the de-esser's detection band, sweep across the vocal, and find where the harshness sits.

Two Approaches

Wideband (single-band). When sibilance exceeds threshold, the entire signal ducks. Simpler, but it audibly dips the whole voice on every "S," which can sound un-natural.

Multiband / split-band. When sibilance exceeds threshold, only the sibilant band ducks. The rest of the voice stays at full level. Sounds more transparent. Almost every modern de-esser does this.

Setting It Up

1. Find the sibilant frequency. Sweep a narrow EQ band on the vocal until the harshness becomes obvious. Note that frequency.
2. Set the de-esser's detection band to centre on that frequency.
3. Lower the threshold until you see 2–4 dB of gain reduction on the loudest "S" sounds.
4. Listen in context. A de-esser that sounds great solo can dull the vocal in the mix.

Common Mistakes

Over-de-essing. If you reduce sibilance below the level of normal speech, the vocal sounds lispy. 2–4 dB on the loudest moments is usually enough.

Wrong band. A de-esser tuned to 5 kHz on a singer whose sibilance is at 8 kHz won't do anything useful and may dull other parts of the vocal.

De-essing too late in the chain. If you compress, distort, or saturate after the de-esser, those processes can re-introduce sibilance. De-ess after the colourful processing, before the limiter.

Plugin Recommendations

Further Reading

• FabFilter Pro-DS Official Page
• Soothe2 Official Page