Rys Up Audio
← Back to blog

How to De Ess Vocals Without Killing the Air

Jordan Rys

Quick answer: To de ess vocals without killing the air, loop the worst S, SH, and T sounds, audition the detector, search the upper range until those consonants jump out, then lower the threshold only until the harsh hits trigger reduction. Cap the range gently, bypass in the full mix, and stop before the singer sounds lispy.

Disclosure: Rys Up Audio makes RysUpDS, which appears in this guide. The workflow still works with any de esser that lets you hear the detector, choose a frequency range, set a threshold, and limit the amount of reduction.

RysUpDS de esser interface with threshold range audition and detector controls

Sibilance or constant harshness

A de esser is built for consonants that leap out for a moment. It should react when the singer says words like sweet, shine, choice, or time, then relax during the vowel and body of the voice. If the whole performance feels sharp from beginning to end, the problem may be a steady upper mid resonance instead of sibilance.

That distinction matters. Crushing every S will not fix a constant bite around the upper mids. It only turns a harsh vocal into a harsh vocal with missing consonants. Use a dynamic EQ or a resonance tool for a tone that stays painful across whole words. Use a de esser when the problem arrives and leaves with specific syllables.

Start by listening in the full mix. The solo button can make normal consonants feel scarier than they really are. If the S sounds only jump out after bright EQ, heavy compression, or saturation, keep those stages active while you set the de esser.

Find the harsh S frequency first

Do not begin by dragging the threshold down. Begin with the detector. Loop the harshest phrase in the song, turn on Audition or sidechain listen, and sweep through the broad 4 to 10 kHz area. You are listening for the setting where the offending consonant becomes the clearest thing in the detector signal.

The exact frequency changes with the singer, microphone, angle, room, and processing. A chart that says male or female vocals always live at one number is a clue, not a rule. One singer may spit energy around 5 kHz. Another may be painful closer to 8 or 9 kHz. Find the event you actually need to control.

Narrow the detector enough to ignore unrelated cymbals, breaths, and vocal body, but not so much that the de esser misses SH, CH, or T sounds that sit beside the strongest S peak. The goal is a detector that hears the problem more clearly than it hears the song.

How to de ess vocals in six moves

  1. Loop the worst phrase: Pick a line with several different consonants, not one isolated S.
  2. Choose the vocal mode: Use the normal single vocal or stereo setting before reaching for a complex mode.
  3. Audition the detector: Sweep until the harsh consonants become obvious and the vocal body becomes quieter.
  4. Lower the threshold: Stop when the problem syllables trigger and ordinary vowels mostly stay untouched.
  5. Limit the range: Use the least maximum reduction that stops the consonant from jumping forward.
  6. Bypass in the mix: Match level, switch the processor out and in, and check whether the words remain natural.

That order keeps threshold and range from doing the detector's job. If the de esser reacts to the wrong sound, more reduction makes the mistake louder. Fix detection first.

Use this RysUpDS starting workflow

Reset RysUpDS, choose Single Vocal, Stereo, and Wide Band for a clean lead. Turn on Audition and move the detector band until the aggressive consonants dominate what you hear. The reset detector spans roughly 3 to 9 kHz, so it is already near the normal search area, but the actual vocal decides where it ends.

Leave lookahead at its 5 ms reset value. Lower Threshold only until the harsh syllables cross it. Keep Range light at first. If the entire word ducks or the vocal loses air, switch to Split Band before adding more processing. Split Band keeps the low body steadier while the high branch moves.

Raise lookahead only if the sharp leading edge slips through before reduction begins. More lookahead is not automatically better. It changes timing and latency, so use the smallest move that catches the consonant. Current RysUpDS 1.3.12 provides Threshold, Range, Single Vocal and Allround modes, Wide Band and Split Band, Stereo, Mid, and Side processing, Audition, detector filtering, and up to 15 ms of lookahead.

Place the de esser where the problem exists

A safe modern chain is corrective EQ, compression, de essing, tone, then reverb and delay. That lets the de esser hear the brightness and level changes created by the main shaping stages, and it prevents an effect return from repeating the harsh consonant.

There is one important exception. If a huge S burst is driving the compressor and making the whole vocal pump, use a light first pass before compression. Then use a second gentle pass later only if the compressor or bright EQ exposes new sibilance. Two processors are not a flex. Each one needs a separate job.

Correct tone
Remove rumble and obvious mud.
Control level
Shape the main vocal dynamics.
De ess
Catch the exposed consonants.
Add space
Feed a controlled signal to effects.

There is no universal before or after rule because the chain can create the very problem you are fixing. Put the de esser where the sibilance is loudest, then verify the whole chain.

Choose Wide Band or Split Band

Wide Band turns down the full vocal for the instant a consonant triggers. On a clean single lead, that movement can sound natural because the whole voice dips together. The risk is obvious pumping or dullness when the detector fires too long or too hard.

Split Band leaves the low body steady and reduces only the high branch. Use it when Wide Band makes the singer sound like the volume fader is moving, when hi hat bleed keeps opening the detector, or when the source is already dense. Split processing can be more surgical, but it can also sound disconnected if pushed too far.

Do not choose by the word transparent. Match the two modes at a similar perceived reduction and listen to the transition into and out of each consonant. The mode that removes the distraction with less audible processing wins.

Use these de esser settings as a decision table

What you hear Likely cause First move Do not do this
Lisp or missing S Too much range or a threshold that is too low Reduce the maximum range Boost air to hide the damage
Whole word pumps Wide Band is moving too much vocal Try Split Band or less range Lower the threshold again
No reduction Wrong detector band or high threshold Audition and sweep again Guess from a vocal chart
Leading edge escapes Detector reacts after the sharp transient Add a small amount of lookahead Max every timing control
Harshness remains The issue is constant resonance, not sibilance Use dynamic EQ or resonance control Erase every consonant

Fix lispy or dull vocals fast

A lisp means the processor is changing intelligibility, not just comfort. Raise the threshold or reduce Range until the natural edge of the consonant returns. If the top end still feels dull after reduction relaxes, check whether the detector band is too broad. Do not add a high shelf before proving that the de esser is no longer overworking.

If one or two syllables are brutal while the rest of the take is fine, lower those clips manually before the processor. Manual clip gain gives the de esser a smaller peak to catch and lets you use gentler settings for the whole performance.

If reverb or delay repeats the S sound, de ess the signal before the send. If hi hat bleed triggers the processor, narrow the detector, clean the source, or automate the noisy gap. A de esser cannot know which bright sound belongs to the singer.

Know when another tool wins

Static EQ wins when the vocal is simply too bright all the time. Manual clip gain wins when only a few consonants are bad. Dynamic EQ wins when the painful frequency moves or lasts across vowels. A resonance suppressor such as RysUpSmooth fits moving harshness that is broader than normal sibilance.

De essing should remain a focused decision. It is not a general make the vocal smoother knob. If the source has room reflections, clipping, or constant mic fizz, fix those causes separately.

Run the phrase, mix, and quiet tests

Print three level matched versions of the harshest line: dry, lightly de essed, and intentionally overdone. Hide the names, then compare them in the full instrumental. The light version should stop the S from pulling your attention without changing the singer's identity.

Next, listen quietly. Harsh consonants can dominate at normal level, while an overprocessed vocal loses articulation when the monitor drops. Finally, check headphones and speakers. If one system sounds perfect and the other sounds lispy, the setting is too close to the edge.

Do not judge only the reduction meter. A large number can be harmless on one short event and obvious on another. The word, the mix, and the bypass decide.

Use RysUpDS for the complete workflow

RysUpDS puts the detector, Audition, Threshold, Range, Wide Band, Split Band, stereo focus, and lookahead controls used in this guide in one interface. Version 1.3.12 is the current signed release for macOS and Windows in AU, VST3, and AAX through RysUpHub.

Own RysUpDS for $49.99, or use RysUpSuite for the current paid plugin catalog plus future paid releases under the current entitlement policy. Browse the full plugin lineup before choosing.

Keep the air and lose the sting

Find the problem before reducing it. Audition the detector, let only the harsh consonants cross the threshold, cap the range, and judge the result in the mix. If the singer starts sounding smaller, darker, or less intelligible, back up.

Use the vocal de esser buyer guide when you still need to choose a paid tool. Use the free de esser guide when the budget is zero. Use the complete vocal mixing guide when you need the rest of the chain. This page owns the hands on setup and troubleshooting job.

Sources checked