The floor, lowered — continuously.

A real broadband denoiser for hiss, room tone and hum — not a gate, not an expander with a marketing name. ROGERIAN DIA's noise reduction tracks the noise floor continuously using minimum-statistics estimation: it watches roughly a second of spectral history and follows the floor as it drifts — air-conditioning that cycles, room tone that changes across a scene — with no noise print and no learn step required. Set the amount, set the sensitivity, and it simply keeps up.

• · Dual Engines — Zero-latency multiband path for tracking and picture work. High-resolution spectral tier (MMSE-LSA with speech-presence steering) for surgical dialogue restoration without watery artifacts.
• · Transparent Routing — The stage is routable in the chain and bit-transparent completely off. Amount and sensitivity controls do exactly what they claim.

Cuts that exist only while the problem does.

Every room rings somewhere. Every voice, some days, bites. The resonance stage is a dynamic, cut-only spectral processor that finds narrow resonances — boxiness, harshness, ring, proximity build-up — and attenuates them only while they sound, at the depth they demand, then gets out of the way. Nothing is notched permanently.

• · Soft / Hard — Adaptive tracking relative to voice context, or absolute thresholds for surgical work.
• · Focus & Detail — Persistence weighting treats sustained rings but spares transient consonants.
• · Display Steering — Five draggable bands over a live spectrum. Drag up where rooms ring, pull down with Depth, modify the response via REFINE (scale, shift, tilt).
• · Delta Listen — Solo exactly what's being removed. The fastest honesty check in the plugin.
• · Reconstruction Paths — Switch click-free between Zero (near-zero latency), Linear (phase-perfect offline pass), and Natural (minimum-phase).

Seven bands. Three characters. Zero latency.

A seven-band parametric built on analog-matched filters: the bell you see at 12 kHz is the bell you get — magnitude matched to the analog prototype all the way up, with none of the high-frequency cramping of textbook digital EQ. Minimum-phase, zero-latency, and drawn over a real-time pre/post analyzer.

• · Full Shape Set — Bells, 12/24/48 dB-per-octave cuts, notches, tilts, band-passes.
• · Per-Band Dynamic EQ — Set a dynamic range and the band eases in only when the signal crosses its self-adjusting threshold. A de-esser, a mud ducker, a presence lift — each is one band, one gesture.
• · CHARACTER System — NEUTRAL (standard bell), COOL (flat-topped plateau for broad tonal moves without a hot spot), and SHARP (focused tight peak for bite without smear).

A voice leveler with its color on a separate knob.

Feed-forward and program-dependent, built around a smooth decoupled detector to hold a read at a consistent level without pumping or dulling consonants.

• · VOICE — Clean, Warm, or Punch dynamic temperaments.
• · CHARACTER — Separate color section. Drive controls Tube, Diode, or Bright saturations independently from gain reduction. Pre/Post routing enabled.
• · Continuous Detection — Peak to RMS blending, program-dependent auto-release, sidechain high-pass (20–400 Hz), and a hard RANGE floor. Lookahead runs up to 3.5 ms on a fixed automation-safe budget.
• · Metering — Live transfer curve alongside a scrolling GR meter so you see the last several seconds of leveling decisions.

The mouth, minus the mouth noises.

Clicks, ticks, lip smacks — the tiny transients that survive every other stage and cost editors hours of clip-gain surgery. Detection finds the click; repair is model-based: an autoregressive engine learns the voice from the audio around the defect and resynthesizes what should have been there, with a hybrid fallback interpolation that guarantees a clean join. Restoration-suite technique running inside the strip in real time.

• · Click Size — Scope the repair from single-sample ticks to fatter mouth events.
• · Sample-Accurate Reporting — Latency is reported exactly to the host, even when switching click sizes dynamically.