Four Technical Facts Revealed
M-De-Esser 2’s Architecture: Four Technical Facts Revealed
De-essing typically starts with a simple move: detect the harshness and reduce the high-frequency region around it. It works. It also treats a vocal S, a hat, a guitar, and a mix bus as if they were the same problem, and it usually takes the air with it.
M-De-Esser 2 is built on a different method, and for a specific seat in the studio. It is the quality-first de-esser in the Techivation line: maximum transparency and precision for mixing and mastering. That resolution costs latency, so it is not the plugin for tracking or live work. T-De-Esser Pro Mk2 is the low-latency processor for that. AI-De-Esser is the short path, where the models set the process so you can keep moving. M-De-Esser 2 is the one you reach for when the decisions should be yours and the top end has to survive them.
Four parts of its architecture explain how.
A conventional de-esser compresses a region. When it acts, the whole region moves, taking useful brightness down along with the problem. M-De-Esser 2 treats the signal as many narrow frequency regions and reduces only the ones that are actually offending. The tone around the problem stays where it was.
Two details matter for quality. Its frequency resolution scales with the session sample rate, so a 96 kHz mix gets the same precision as a 48 kHz one rather than half of it. And the dry path and the difference monitor are both delay-matched to the processing, so Mix and Diff are phase-honest instead of smeared. What you hear at 50% is genuinely half the processing, not a comb filter.
Harshness is not a fixed frequency. A male S, a female T, a snare, a distorted guitar and a bright mix do not live in the same place, and a single vocal does not stay put from word to word.
Adaptive targeting measures where the energy is concentrated right now, inside the range you allow, and processes that moving region. The frequency limits, anywhere from 800 Hz to 20 kHz, are a fence, not a band. You say where it may look; it decides where the problem is. Both ends of that fence are real, so you can protect the extreme top, and tightening it does not weaken the reduction inside it. Adaptive Range can also be switched off when you already know the band and want the processing to stay there.
Focus governs how the plugin behaves inside that fence. It sets how narrowly the engine locks onto peaks, how firmly reduction begins, and how the edges of the processed region behave. One knob controls what are really several decisions. Low Focus is wide and soft, for a drum bus, a guitar, or a mix that should settle without a notch. High Focus is narrow and decisive, for a lead vocal where only the S should move.
The fence has a third dimension: the processing can run in mid/side, and suppression can be weighted between channels. De-ess the centre of a mix without touching the sides, or lean the reduction toward the channel that actually has the problem.
Generally, de-essers ask for a threshold at an absolute level. That holds until the singer leans in, the clip gain changes, or the same setting moves onto a bus. Then the plugin is either lazy or overworked.
M-De-Esser 2 draws no flat line across the highs. It builds a contour from the incoming signal and compares each region against that contour, relative to the overall peak. Sensitivity is not “how loud is the S in the DAW.” It is “how far does this energy stick out from the rest of the sound.”
That is why one Sensitivity setting can hold through a quiet verse and a loud chorus, and why the plugin works on far more than dry vocals. A bright mix already has a high-frequency shape; a flat threshold shaves that shape down, while a relative one takes the spikes and leaves the shape intact. if Sensitivity is so high that mids and formants start moving, you have gone past smoothing into darkening.
The standing complaint about de-essers is that they make things dull. Removing energy is the only tool they have.
M-De-Esser 2 has a second tool. Air is a dynamic harmonic stage that works on the same frequency limits you set for the engine, following the level of the material rather than applying a fixed amount. Because it generates harmonics, it runs at up to 8× internal resolution so the added content stays clean instead of folding back into the audible band as grit. That is what the Standard-to-Ultra quality setting controls.
Air is also its own control, not a by-product of how hard you're suppressing. Strength decides how much of the detected harshness comes down; Air decides how much presence goes back into that same range. Heavy suppression with a touch of Air is a clinical clean-up. Light suppression with generous Air softens an edge while opening the top. Suppress the spike, return the presence.
Listen for the harshness coming down while the phrase keeps moving. No grab, no hole in the word - just a cleaner top.
M-De-Esser 2 is not trying to be every de-esser. It is the one for when the top end has to stay open, the reduction has to stay exactly where the problem is, and what matters is the mix, not the input.
![M-EQ Review By Czar [Video]](https://cdn.sanity.io/images/9uzzx98j/production/341fd252af5cfdce3815b27290e6ccf6b1185b98-1280x720.jpg)