In this guide
Listen to complete words before treating an S sound. A small level edit can help an isolated consonant; a de-esser dynamically controls sibilance across a longer performance. Compare at matched playback levels and stop before speech becomes indistinct.
Sibilance is the energy in sounds such as S, SH and CH. Those consonants carry meaning, but a particularly sharp one can distract in headphones or a voiceover. The goal is to make the sound sit comfortably in the phrase while keeping the word intelligible, not to erase it.
First compare the recording with its original at a comfortable, similar playback level. If every consonant sounds sharp only on one device, also listen on another. If the problem is a low thump on P or B, use the plosive guide; it is a different kind of event.
Choose a method for the pattern
| What you hear | Starting point | Check afterward |
|---|---|---|
| One distracting S between otherwise natural words | A small, smooth local volume dip on a copy. | The whole word still sounds clear and the transition is inaudible. |
| Harsh sibilance throughout a performance | Try a dedicated de-esser in an audio editor. | Different words, quiet phrases and the complete mix remain natural. |
| Steady hiss in the pauses | Compare a noise-reduction method using an appropriate noise sample. | Wanted consonants and room sound are not damaged. |
| A low-frequency puff on a P or B | Inspect that event separately as a possible plosive. | The consonant and the body of the voice remain intact. |
Make a small local edit
- Save a working copy. Identify the exact word by listening, then zoom in. A waveform peak alone does not identify an S sound.
- Mark the consonant and its context. Include enough of the preceding and following speech to hear how it joins the word. Avoid cutting the consonant away from the vowel.
- Use a gradual level dip. With clip gain or volume automation, lower the troublesome interval slightly and smooth both edges. Audacity's Envelope Tool manual describes volume control points. Use outer points to hold the surrounding level and inner points for the dip.
- Listen to the whole phrase. Compare with the untouched original at a similar level. If the speaker sounds as though they have a lisp, the word is less clear, or the room tone dips noticeably, reduce or undo the edit.
A volume dip lowers everything in the selected interval, including any voice or background underneath. It is not frequency-selective. Keep the original duration when the voice must stay aligned with picture or another track; deleting time moves what follows.
If you use an Audacity envelope, check it after applying later effects. The Envelope Tool manual warns that some effects can remove the envelope and restore the original level. Keep a working copy, and compare the exported phrase with your approved edit.
Try a de-esser for recurring harshness
A de-esser is designed to reduce sibilance dynamically as it occurs. Controls and detection methods vary by processor. Start with its documented defaults on a duplicate track, play several representative phrases, and increase reduction gradually only until the sharp sounds no longer distract. The iZotope vocal-editing guide discusses both manual and automatic de-essing and cautions against removing too much.
Check soft and loud words, not only the worst S. A processor can react differently across a performance, and aggressive reduction can make consonants dull or indistinct. If the plugin offers a difference or monitoring mode, consult its own instructions and return to normal playback for the final decision. Keep a version without the effect for comparison.
Know what the other tools do
A noise gate responds to level rather than recognizing consonants. It may open for a loud S and can reduce a quiet word; the free noise gate is for comparing level-based reduction, not a de-esser. A high-pass filter reduces low frequencies and is not the same operation as controlling harsh high-frequency consonants. A static equalizer changes its selected frequencies whenever they are present, including wanted voice detail.
Source isolation estimates a described sound from a mixture; it is not a dedicated sibilance repair. If another sound competes with the voice, compare approaches in the cleanup-method guide. A quieter result alone does not establish that the speech improved.
Prevent and review on the next take
When recording again is possible, make a brief test and experiment with speaking slightly across the microphone instead of directly into it. Listen before committing: distance and angle can also change voice tone and room sound. Avoid treating a single microphone position as a universal fix.
After editing, listen in the intended mix and on another playback system. Check that names and consonants remain clear, and review the final export after any further compression or encoding. For an intrusive inhale, use the separate breath-editing workflow; for dialogue returned to picture, follow the video sync checks.