Production guides · Meme Plugins · 27 October 2026
De-ess a vocal by finding the sibilant events, tuning the detector to them and applying only the reduction needed when they occur. Check ordinary words and breaths as well as the sharp consonants. A vocal that becomes dull or starts to lisp usually needs a more selective or smaller change.
De-essing is a specific dynamics task. If the words are hidden by the arrangement, start with the vocal-clarity workflow. More sibilance control does not fix every intelligibility problem.
1. Mark the distracting consonants
Loop a phrase containing the sharp event and several ordinary syllables. Compare the raw vocal with its processed chain to find whether EQ, compression or another stage makes the issue more obvious. Keep the source and a bypassed version.
Do not assign one frequency to every singer or microphone. Search for the actual event, using a detector-listen function when the device provides one, then return to the full signal.
2. Adjust detection and reduction separately
FabFilter’s de-esser controls distinguish the trigger’s frequency region, threshold and reduction range. In that device, listening to the trigger helps identify what activates processing. Your device may offer a different detection method or naming.
First make the detector respond to the relevant consonants without repeatedly catching normal vowels. Then set a restrained reduction range. If processing is active through nearly the whole phrase, reconsider the trigger and threshold rather than only turning the amount down.
3. Compare what gets attenuated
FabFilter’s processing reference distinguishes full-band gain reduction during detected events from split-band reduction of higher frequencies. This is a choice about what changes after detection; it is separate from selecting the trigger region.
Compare the available modes on the same line. Listen for a whole-word level dip in one case and an unnatural change of consonant tone in the other. Neither mode is universally more transparent.
Example: one sharp phrase, two treatments
Imagine an original spoken line, “Soft stars settle,” with an exaggerated first sibilant and otherwise useful brightness. This is a hypothetical comparison, not an audio example we recorded or measured.
- Permanent-cut version: reduce the high-frequency region with ordinary EQ through the entire line. Note whether the unaffected words and breath lose useful detail.
- Event-controlled version: restore the baseline and tune a de-esser so it responds mainly to the excessive sibilant. Adjust the amount until that event fits the rest.
- Matched comparison: check the phrase in the backing track at comparable playback level. Listen for clear pronunciation, normal vowels and a stable vocal position.
- Exception: if only one consonant remains excessive, compare a small local edit rather than making the whole detector more aggressive.
Write separate notes for the sharp event and the unchanged words. A smoother sibilant is not a useful result if the remaining vocal loses its character.
4. Check the chain and another section
De-essing before or after compression can produce different results because the signal reaching each detector changes. Recheck settings after moving a stage. Use deliberate rides and compression for broader phrase balance.
Check a louder section and the reopened export. Keep the least intrusive version that solves the specific distraction.
Should every s sound be reduced?
No. Sibilants are part of speech. Reduce distracting excess while preserving intelligibility.
Is harshness always sibilance?
No. For a sharp guitar, cymbal or wider tonal problem, use the harsh-mix diagnosis.