De-esser

What it is and how it works

A de-esser is a compressor that reacts only to sibilance (s, z, sh, f, ch) and hard consonants. The detector listens to a filtered version of the signal (band-pass or high-pass on the sibilant zone): when the energy in that zone exceeds the threshold, the de-esser lowers the signal. Two ways of reducing:

• Split-band (band only): lowers only the sibilant frequencies; the body of the voice stays intact. This is the most used mode.
• Wideband (whole signal): lowers the entire voice during the sibilant. More natural on light sibilants, but at high values it makes the voice “stumble”.

The de-esser is needed because compression, presence EQ, saturation and air (10–12 kHz shelf) increase sibilance.

Where sibilance lives

Find the frequency with a sweep: narrow band +10 dB, sweep between 4 and 11 kHz while the voice says “s”: the spot where the “s” hurts most is the center.

Parameters one by one

• Frequency: center or corner point of the detector's filter (see table).
• Mode: split-band or wideband; detector filter as a bell or high-pass (the HP catches all the sibilants above a point, the bell is more precise).
• Threshold: lower it until the GR appears only on the sibilants. If the reduction also appears on the vowels, it is too low.
• Range / Ratio: how much is reduced at most. Typically 3–6 dB, up to 8–10 dB on very sibilant voices. Beyond that, the voice sounds “lispy” (the s turns into “th”).
• Attack/Release: often automatic. Attack 0.5–2 ms, release 30–80 ms (a sibilant lasts 50–200 ms).
• Lookahead: 1–5 ms to catch the attack of the “s” without letting it slip through.
• Listen / Audition: listen only to what the de-esser removes. It should sound like a series of “s” sounds and nothing else.
• Stereo link / M-S: on stereo backing vocals or vocal buses, 100% link or processing on the Mid.

Tools

• FL Studio stock: Maximus (use the High band, with the crossover at 4–6 kHz, fast compression and the other bands bypassed; the manual lists this among its uses), Fruity Multiband Compressor (same logic), automated Fruity Parametric EQ 2. There is no dedicated stock de-esser.
• Third party: FabFilter Pro-DS (Single Vocal / Allround modes, lookahead, split-band), Waves Sibilance and DeEsser, Pro-Q in dynamic mode, TDR Nova (free), oeksound Soothe 2 (dynamic resonance suppression, also excellent for harshness), iZotope Nectar/RX.
• Manual: lowering each “s” by hand with volume automation or clip gain (in Edison or in the Playlist by splitting the audio). The most transparent method for very exposed vocals.

Where it goes in the chain

• After compression (which raises the sibilants) and before the air EQ is the most common position.
• Two de-essers: a light one before the compressor (so the compressor doesn't react to the “s” sounds) and one after, each with 2–3 dB.
• On the reverb and delay send: a de-esser at the effect's input (or a send from a track already de-essed) avoids “s” sounds bouncing in the reverb; you can be more aggressive (6–10 dB).
• On the vocal bus to even out doubles and backing vocals: the overlapping “s” sounds of several voices add up and become very obvious. In the doubles, the sibilants are often simply cut (left only in the lead vocal).

Starting settings

By genre

• Pop/K-pop: very bright vocals (air +3/+5 dB) require decisive de-essing, often in two stages plus Soothe.
• Rap/trap: compressed and saturated vocal: de-esser after saturation and compression.
• R&B: falsettos and overlapping backing vocals: strong de-essing on the backing vocals.
• Rock: vocal often less bright, light de-essing; watch out for cymbals bleeding into the microphone.
• Acoustic/jazz: minimal or manual de-essing.

Common mistakes

• Excessive range: lispy, dark vocal.
• Threshold set on a single phrase: elsewhere it doesn't act or acts too much (check the whole track).
• Frequency too low (3–4 kHz): removes intelligibility, not the sibilants.
• Forgetting reverbs and delays: the dry vocal is clean, but the “s” sounds remain in the reverb.
• Using a static EQ for sibilance: you darken the whole vocal.

How to hear it

• Turn on Listen and hear only the removed signal: just “s” sounds, no vowels.
• Listen at high volume on headphones (sibilants are more obvious) and then on the phone.
• Compare on a word with lots of “s” sounds: the “s” should stay clear but not sting.