De-esser
A de-esser tames the harsh s, sh, ch and t sounds — called sibilance — that bright microphones and close-up speaking exaggerate. It listens to one slice of the spectrum, and only when a sharp "ess" pokes out does it briefly turn down the top end of the sound. The rest of your voice — its body and warmth — is never touched.
Reach for it when:
- every s on your stream or recording hisses or stings, especially on headphones;
- you've boosted the treble (or added a Compressor) for clarity and the s sounds came up with it;
- a guest or co-host on Discord sounds spitty on a bright mic.
FX inserts can be added and set up on Free, but they only affect the sound with Premium.
Adding it
On a channel's FX tab, click + Node ▾ → Dynamics → De-esser. It drops into the chain just after the selected node. To edit it, pin one of the dock's two panels to it with ⇅, or set a panel to ⟳ Follow FX tab selection and click the node.
In Simple mode, add it from the Effects block's ⚙ instead — see Simple mode.
Where it sits in the chain matters: put it after the Compressor. A compressor lifts quiet detail, sibilance included, so de-essing after it catches the s sounds at their loudest. Rewiring the chain explains how to move it.
Reading the panel
The panel has two graphs side by side.
The left graph shows where the de-esser is listening — a hump centred on the FREQ you set, across the range from 200 Hz to 16 kHz. The hump is fairly broad on purpose: sibilance isn't one exact pitch, so the detector hears a band around your chosen frequency rather than a single note.
The right graph shows how hard it reacts once that band gets loud. Level coming in runs left to right, level going out runs bottom to top. Below the threshold dot nothing happens (a straight diagonal line); past it the line bends down according to RATIO. The dashed line is the RANGE floor — the de-esser never turns the top end down further than that.
While the engine is running, the right graph comes alive: a bar along the bottom shows the incoming level, and the GR strip on the right edge shows how many dB the de-esser is taking off right now (the number under it). Say "sister, sister" and you should see it kick in on the s and drop back to 0 in between.
SPEC in the panel header (on by default) lays a live spectrum analyzer behind the left graph, so you can see where your s sounds peak and slide FREQ onto them.
Need more room? ⛶ in the header maximizes the panel to full dock height.
Controls
| Knob | Range | Default | What it does |
|---|---|---|---|
| FREQ | 1,000…12,000 Hz | 6,000 | Where the harshness lives. It centres the listening band, and it's the point above which the sound gets turned down when an s is caught — everything below it is left alone. Deeper voices usually sit around 4–6 kHz, brighter voices around 6–9 kHz. |
| THRESH | −40…0 dB | −24 | How loud an s has to be before the de-esser reacts. Lower = more de-essing, because quieter s sounds start to trigger it too. |
| RATIO | 1…20 : 1 | 6 | How hard a caught s is pushed down once it's past the threshold. 1:1 does nothing; higher is firmer. |
| RANGE | 0…18 dB | 12 | The most it will ever turn the top end down, however loud the s. A safety limit that stops it from swallowing sibilance completely. |
| ATTACK | 0.1…10 ms | 0.4 | How fast it grabs an s. Keep it very short — sibilance is over in a flash. This also sets how quickly the de-esser notices an s, so a slow attack makes it miss the start of one, not just react gently. |
| RELEASE | 1…50 ms | 4 | How fast the top end comes back after the s ends. Short keeps the next word bright; too short can make a flicker on long sss sounds. |
Left and right are always treated together: if either side has a loud s, both are turned down equally, so the stereo image never shifts.
A good starting point
- Add the De-esser after the Compressor, leave every knob at its default, and speak normally — include a few sibilant phrases ("she sells sea shells", "this is so simple").
- With SPEC on, watch the left graph as you say sss: a peak lights up somewhere between 4 and 9 kHz. Slide FREQ so the hump sits on it.
- Lower THRESH until the GR strip reacts to the s sounds, but stays at or near 0 on ordinary vowels.
- Listen with the node ON and OFF. The s should sound softer, not gone. If it starts to sound like a lisp, back THRESH up a few dB or lower RANGE.
A good de-esser is one you don't notice. Aim for 3–6 dB on the loudest s sounds; if GR is showing 10 dB or more all the time, it's working too hard.
If it sounds wrong
| You hear… | Try… |
|---|---|
| A lisp — like the speaker lost a tooth | Raise THRESH, lower RATIO or RANGE. |
| The whole voice goes dull or muffled whenever you talk | FREQ is too low, so it's turning down too much of the top end — raise it, and check THRESH isn't triggering on normal speech. |
| s sounds still sting | FREQ isn't on the harsh spot — use SPEC to find it; then lower THRESH. |
| The start of each s slips through | Shorten ATTACK. |
| A flutter or pumping on long sss sounds | Lengthen RELEASE a little. |
Presets, bypass and reset
- PRESET ▾ in the panel header applies a saved setup, or saves the current one under a name. Once one is applied the header shows Preset: name — see Knowing what preset is applied. Ready-made de-esser setups can be found in the Workshop — Workshop ▸ Browse, filter chip De-esser.
- ON / OFF bypasses just this node — the quickest way to hear whether it's helping.
- RESET returns every knob to its default.
Related
- Equalizer — an EQ cut on the treble dulls the voice all the time; a de-esser only acts the instant an s appears. Use the EQ for tone, the de-esser for sibilance.
- Compressor — the usual neighbour; compress first, then de-ess.
- Effects Rack — adding, ordering and rewiring nodes.