De-Esser
Chamfer De-Esser
Eases the edge. Rides the performance. Shows every ess.
A de-esser that reshapes each ess on its own, tracks the performance level so the de-essing does not change when the singer does, and draws its actual threshold on the display. Click any ess and see exactly what was done to it.
Signal path
Why it exists
A threshold set for the verse is wrong for the chorus, and a de-esser deep enough to fix the ess usually takes the vowel with it. Chamfer's threshold is relative and level-tracked: turn the singer up 18 dB or down 18 dB and the delivered depth moves by 0.5 dB or less. Each ess is handled on its own, reshaped spectrally rather than ducked, and at matched delivered depth on the same takes held-vowel body moves 0.00 dB on average. A frozen 10 dB floor on ess over body means it cannot be driven into a lisp. And because the display's input tap is latency-aligned to the output tap, the ride line you see is the detector's own reference and the removed-energy shade cannot show reduction that did not happen.
The glass
The ride is the threshold. Every ess is a card.
01 · Ride
The ride line is the actual threshold, live.
A real take through the live processor. The drawn ride is the detector's tracked reference, K-weighted over 250 ms and calibrated on real material so esses visibly cross it when they trigger. Turn the singer up 18 dB or drop them 18 dB and the delivered depth moves by 0.5 dB or less at the command level, 0.75 dB or less measured end to end on rendered audio.
02 · Inspect
Click any ess and see what was done to it.
The ess strip records every event; click one and the inspect card freezes its pre and post spectra captured at the event's deepest hop, with the removed energy shaded and the centroid marked. Identically constructed events come out identical: the residual between them after processing sits below minus 120 dBFS on every style and mode.
03 · Live
4 ms when mixing. Zero when tracking.
Precise mode reports and measures exactly 191 samples at 48 kHz. Live mode reports and measures zero, for tracking. Style changes glide gains on one engine; they never relatch or click.
04 · Styles
Three styles. One lisp floor.
Smooth, Split and Surgical run on one engine, and every one of them holds the frozen 10 dB floor on ess over body. Lisp floor breaches across the 8-take battery and all 23 preset audition renders: zero. The dulling metric reads 0.00 dB mean across every audition render.
Where it earns its keep
Built for the jobs you actually have.
Lead vocals
Set it on the verse and forget the chorus: the relative threshold tracks the performance level, so the delivered depth holds across an 18 dB swing either way.
Backing vocals and stacks
Per-event reshaping keeps every layer's vowels intact while the esses come down, with Split and Smooth styles for stacks that want less surgery.
Dialogue and voice-over
The click, breath and bright-vowel traps are standing tests, so plosives and breaths are not mistaken for esses; detector precision on the labeled battery is 1.000 at defaults.
Hats, overheads and acoustic guitar
Starting points named for sources beyond the voice, each measured over its matched real take before it shipped: real reduction, zero lisp breaches, no dulling.
What's in the unit
Real pixels, not promises.
The ride line
The detector's tracked reference, drawn live over the spectrum, with the removed-energy shade from latency-aligned input and output taps.
The inspect card
Pre and post spectra frozen at the event's deepest hop, the removed energy shaded, the centroid marked, the depth and confidence printed.
Precise or Live
191 samples at 48 kHz for mixing, zero for tracking, with threshold, detect and ballistics in the drawer.
Smooth, Split, Surgical
Three styles on one engine. Style changes glide gains and never click, and every style holds the same frozen lisp floor.
Measured, not promised
| Claim | The measurement |
|---|---|
| Turn the singer up 18 dB or down 18 dB; the de-essing does not change | Delivered depth moves 0.5 dB or less at the command level across a plus or minus 18 dB sweep; 0.75 dB or less measured end to end |
| Deep de-essing that does not dull the voice | At matched delivered depth on the same takes, held-vowel body moves 0.00 dB on average, worst case 1.3 dB |
| It will never lisp you | Frozen 10 dB floor on ess over body; zero breaches across the 8-take battery and all 23 preset audition renders |
| Every ess handled on its own | Residual between identically constructed events after processing below minus 120 dBFS on every style and mode |
| No chatter | At most 4 gain flips per cycle allowed; measured median episode 46 ms |
| Plosives, clicks and breaths are not esses | Click, breath and bright-vowel traps are standing tests; detector precision 1.000 at defaults on the labeled battery |
Everything in the box
- Relative, level-tracked threshold: depth moves 0.5 dB or less across a plus or minus 18 dB program sweep
- Absolute threshold mode when you want one
- Per-event spectral reshaping: identical events come out identical, below minus 120 dBFS residual
- Frozen 10 dB floor on ess over body, enforced where gain is applied: zero lisp breaches measured
- Held-vowel body moves 0.00 dB on average at matched delivered depth
- The ride line on the display is the detector's tracked reference, K-weighted over 250 ms
- Display input tap latency-aligned to the output tap; the removed-energy shade cannot lie
- Ess strip: every event recorded, every event clickable
- Inspect card: per-event pre and post spectra at the deepest hop, removed energy shaded, centroid marked
- Precise mode: exactly 191 samples of latency at 48 kHz; Live mode: zero
- Three styles on one engine: Smooth, Split, Surgical; style changes glide, never click
- No chatter: at most 4 gain flips per cycle, measured median episode 46 ms
- Click, breath and bright-vowel traps as standing tests; detector precision 1.000 at defaults
- Listen modes: Band, Voice, Removed
- Stereo, Mid focus and Sides detection with channel link
- A/B snapshots that persist with the session
- 24 starting points named for sources, each measured on its matched real take
- VST3, AU and AAX on Windows 10+ and macOS, Intel and Apple Silicon
Specifications
Questions engineers ask
Does the de-essing change when the singer gets louder?
No. The threshold is relative to the tracked program level. Across a plus or minus 18 dB sweep, delivered depth moves by 0.5 dB or less at the command level and 0.75 dB or less measured end to end.
Can it be driven into a lisp?
No. A frozen 10 dB floor on ess over body is enforced where gain is applied. Zero breaches across the 8-take battery and all 23 preset audition renders.
What latency does it add?
Precise mode: exactly 191 samples at 48 kHz, about 4 ms. Live mode: zero. Both are reported to the host and measured.
Does it work on sources other than vocals?
Yes. The 24 starting points are named for sources, including hats, overheads and steel-string guitar, and every one was measured over its matched real take before it shipped.
Which formats and platforms?
VST3, AU and AAX on Windows 10+ and macOS, Intel and Apple Silicon. That covers Pro Tools, Logic Pro, Ableton Live, Cubase, Studio One, FL Studio, REAPER and every other major host.
Is there a free trial?
Yes. Every released Vitric plugin has a free 3 day trial with every feature available during the trial window. Get the installer and your machine-bound trial license from the downloads page. At the signed expiry, audio processing and generated MIDI lock until you install a purchased license or a valid subscription grant. The plugin can still load so projects and saved settings are not deleted.
More from the line
Chamfer De-Esser
Deep without dull. Proven on the display.
The ride you see is the threshold. The ess you click is the one it touched.