Sibilance and de-essing: how to use a de-esser
Last updated 6 min readTags:MixingVocalsDynamicsEQ
Learn what sibilance is, how to set a de-esser, and where to place it in your vocal chain. Reduce sharp consonants without dulling the voice.
Sibilance is the hiss in consonants such as ‘s’, ‘sh’ and ‘z’. It is a normal part of speech, but it can sound piercing when it is too loud in a recording. A de-esser reduces these sounds briefly, either by turning down the whole vocal or by reducing its high frequencies.
Use one when consonants stand out even though the rest of the vocal sounds balanced. The aim is to keep the words clear without making the singer sound as though they have a lisp.
Is it sibilance or another problem?
Listen to which parts of the words sound unpleasant. Sibilance follows particular consonants; harshness may also affect sustained vowels.
| What you hear | What to check |
|---|---|
| Sharp hissing on ‘s’, ‘sh’ or ‘z’ | Sibilance; try de-essing or manual level edits |
| A persistent sharp edge, including on vowels | Tonal balance; try static or dynamic EQ |
| Low thumps on ‘p’ and ‘b’ | Plosives; edit the affected sounds or filter their low frequencies |
| Rasping or crackling | Possible distortion; check the recording and processing chain |
These problems can occur together. A de-esser may reduce an unpleasant consonant, but it cannot repair a clipped recording. See the guides to equalisation and distortion for those issues.
How a de-esser works
A de-esser detects sibilance and applies a short reduction in level. Many use a filtered side-chain: a separate copy of the signal used for detection. Some also analyse the signal to distinguish sibilance from other sounds.[1]
Two common processing modes are:
- Wideband: reduces the whole signal during a detected sibilant.
- Split-band: reduces the high-frequency portion while leaving lower frequencies alone.[2]
The detection range and the affected frequencies are not necessarily identical. In Pro-DS, for example, the side-chain filters define what triggers processing, while split-band mode reduces the high frequencies above a crossover.[1][2] Check how your plug-in handles these controls.
How to set a de-esser
1. Find the troublesome consonants
Loop a phrase with several obvious examples. Use the de-esser’s side-chain listen or audition control, if available, to hear what its detector receives.[1]
Adjust the detection range until the troublesome consonants are prominent. There is no universal frequency setting: ‘s’ and ‘sh’ can need different treatment, even within one performance. A preset is a starting point, so check it against the recording.
Turn audition off before judging the processed vocal.
2. Compare wideband and split-band modes
Try both on the same phrase. Wideband can sound natural on a single vocal. Split-band preserves the lower part of the voice, but excessive reduction can make consonants sound dull or detached.[2]
Choose the mode that sounds less intrusive at a useful amount of reduction.
3. Adjust the threshold and reduction
The threshold controls when processing begins. Lower it until the de-esser catches the offending consonants. Watch the gain-reduction meter and listen for unwanted changes to the rest of the phrase.
If your plug-in has a Range or Max Reduction control, use it to limit the amount of processing.[1] Increase reduction gradually until the consonants sit comfortably with the vowels. There is no required gain-reduction reading.
If the voice becomes dull or lispy, reduce the amount. If ordinary syllables also trigger processing, revisit the detection range and threshold.
4. Check the whole performance
Listen with the full mix playing, then bypass the de-esser. Keep the vocal’s apparent level comparable. The words should remain clear, with less distracting hiss.
Check quieter verses, louder choruses and phrase endings. If only a few consonants still stand out, edit those individually rather than increasing reduction across the whole song.
If the start of each sibilant escapes processing, try lookahead where available. It gives the detector advance notice of the sound, though it adds latency.[2]
Where to put the de-esser
Place it according to where the problem occurs.
| Position | When to try it | What to check |
|---|---|---|
| Before compression | Strong consonants cause unwanted compressor gain reduction | Whether the compressor responds more evenly |
| After EQ and compression | Sibilance becomes excessive after a bright boost or changes to dynamics | Whether the processed vocal keeps its intended tone |
| Before reverb or delay, on the effects path | Consonants leave a conspicuous hiss in the effect | Whether the effect sounds smoother without changing the dry vocal |
Early de-essing can reduce how strongly sibilants trigger a compressor. The result depends on the compressor’s detector and settings. Later processing may still make those consonants too prominent.
Two gentle stages can be useful when the problem occurs at different points in the chain. Check their combined effect before adding more reduction. Revisit the settings after changes to vocal EQ, compression or saturation.
When another approach works better
For a few isolated consonants, use clip gain or volume automation. Lower each one and listen across the edit boundaries. Short fades can prevent clicks. This takes more time than automatic processing but gives precise control.
For a specific frequency range that becomes excessive, try dynamic EQ. A dynamic band reduces that range as its level rises. A narrow band can treat a localised problem without reducing all the high frequencies.[3]
For a vocal that is consistently too bright, reconsider the EQ. A fixed cut or a smaller high shelf may be more suitable than repeated de-essing.
If another take is possible, test microphone position and angle before recording again. Compare the whole vocal tone as well as its consonants. On stacked vocals, check whether poorly aligned consonants create several separate hisses.
Can you de-ess a finished mix?
Yes, but the processing can also affect cymbals, hi-hats and other bright sounds. Return to the vocal track if you have access to the session.
If only the stereo file is available, try gentle split-band processing. Compare a passage with clear vocal sibilance and one with prominent cymbals. If the drums lose their brightness before the vocal improves enough, reduce the processing or return to the mix. The mastering EQ guide covers broader tonal adjustments.
What a Sonalix report can tell you
A broad high-frequency energy reading cannot identify vocal sibilance on its own. Cymbals, guitars and other instruments contribute to the same range, while brief consonants may have little effect on a whole-track average.
Use the report as a prompt to listen. Find the offending sound before choosing a processor. The 5-band energy guide explains how to interpret broad frequency measurements.
Sources
- FabFilter Pro-DS: Basic controls, detection, side-chain filtering and audition.
- FabFilter Pro-DS: Advanced controls, wideband and split-band processing, crossover behaviour and lookahead.
- FabFilter Pro-Q: Dynamic EQ, level-dependent EQ and vocal sibilance control.
Related pages
Was this page helpful?