EVERWORKS BY FLUGE / SOUND ENGINEERING Fluge
Discord Download
SR48 kHz BIT32f
CPU3.1% CH2.0 FMTAU · VST3 · StandaloneAU · VST3
01 / 11 — THE CLEAN SUITE · CALM EverClean V2

A processor for CALM on stage.

Recover the SOUL after cleanup.

Two-mode live vocal processor. In Zero Latency, EverClean is the primary defeedback and cleanup path: it detects and removes feedback and resonant ringing while reducing noise, bleed and room tail. CALM adds a causal tonal guard. For difficult channels, Studio 21 ms activates the CALM Mamba network and higher-resolution DNS. The focused EverClean engine from CALM as a dedicated one-knob plugin. Its real-time defeedback stage detects and removes feedback and resonant ringing while also cleaning noise, bleed and room tail. Zero reported latency, with no channel-strip complexity. Sub-bass optimisation chain. Aligns, tightens and contours the low end so kicks, toms and bass guitars share the room — without fighting it. Zero-latency intelligent livebus processor for master, matrix, stream and PA buses. ATLAS adds weight, punch, air and depth while GUARD keeps translation, low-end and peak safety under control. Musical algorithmic reverb with a 16-line HyperFDN core, 4D tail modulation, spectral EDR fog and subtle glue so vocals, drums and synths bloom without washing away the source. THE VOID V2 preserves the original Everworks cleanup response in a faster cross-platform engine, measured up to 40.3% lower on Apple Silicon. Turbo extends Void Amount from 100% to 200% for difficult sources without added latency. A dedicated regeneration processor for rebuilding body, continuity and stable detail after cleanup. SOUL keeps the entire workflow to one Regeneration fader and a latency-aligned reconstruction path. Zero-latency adaptive vocal EQ that observes long-term tone, compares it with one of nine tonal populations and applies a bounded 30-band correction while the direct audio remains in the time domain. Intelligent harmonic EQ for 75 instrument and voice profiles across 12 musical contexts, with adaptive target matching, H2-H5 control and measured Live or phase-matched Studio processing. Adaptive low-end instrument with pitch-aware sub synthesis, root reinforcement, resonance, saturation, compression and a dual high-resolution spectrum view. A zero-latency distributed automixer with two global models. Choose EUCLID for spill-aware source focus or CLASSIC for proportional Dugan-style gain sharing; every linked instance follows the same mode while audio remains local.

NO FEEDBACK NO REVERB REAL TIME
↳ Register to Download Free Beta packages in preparation View specs
CALM EverClean V2 · Everworks
CALM EverClean interface with the complete channel strip open EverClean plugin interface
CALM AUTO SRC STUDIO 21ms ZERO LAT EVERCLEAN LIVE READY
SIGNAL ANALYSIS
PROTECTION
CALM
CALM
EVERCLEAN
CLEAN
MAGNETAR V2 interface with dual RTA ATLAS LiveBus interface with dual RTA STARDUST plugin interface THE VOID V2 interface with Matter, Turbo, Body, Consistency and Matrix controls SOUL 0.1.0 regeneration plugin interface D.N.A VOCALS intelligent vocal EQ interface D.N.A ULTIMATE intelligent harmonic EQ interface LEVIATHAN adaptive low-end instrument interface QUORUM 0.3 linked automixer with global EUCLID and CLASSIC modes
EVERWORKS STARDUST Nimbus Vocal EDR TAIL FOG
SPACE ENGINE
MIX38%
PRE28 ms
SIZE64%
DECAY CORE
DECAY4.8 s
DIFFUSIONTEXTURESMOOTHGLUE
TAIL TONE
LOW CUT120 Hz
HIGH CUT16 kHz
WIDTH108%
/ 01Overview

Built for the people.

CALM EverClean has two distinct workflows. In Zero Latency, EverClean is the main defeedback and cleanup engine, removing feedback and resonant ringing while CALM adds causal narrow-band protection. Studio 21 ms is the escalation path with CALM's Mamba network and higher-resolution DNS. EverClean is the dedicated one-knob defeedback and source-cleaning member of The Clean Suite. It suppresses feedback and resonant ringing while reducing noise, bleed and room tail, with zero reported latency. MAGNETAR V2 is part of The Mastering Suite: a five-macro master-bus enhancer with a dual 8192-point RTA, A/B snapshots, gain match and delta monitoring. ATLAS LiveBus is part of The Mastering Suite: a zero-latency intelligent bus processor for live masters, matrices, streams and PA outputs. STARDUST is the third Everworks suite plugin: a musical algorithmic reverb built around a 16-line HyperFDN, tesseract-style delay motion and dynamic Householder feedback. THE VOID V2 preserves the original Everworks model and its familiar cleanup response while reducing live CPU, improving cross-platform performance and adding Turbo for difficult sources. SOUL is the regeneration member of The Clean Suite. Insert it after cleanup and raise one fader to recover body, continuity and stable detail without reopening the complete processing chain. D.N.A VOCALS is an adaptive vocal EQ that compares the stable character of a source with nine tonal populations and applies the result through a causal 30-band filter bank. D.N.A ULTIMATE extends intelligent target matching to voices, instruments, drums and complete mixes, then adds a controllable H2-H5 harmonic layer without hiding the spectrum. LEVIATHAN is an independent adaptive low-end instrument: six immediate macros open into a deep Track, Sub, Tone, Dynamics and Output architecture without adding host latency. QUORUM is a distributed automixer with EUCLID and CLASSIC gain-sharing models. One global selection updates every QUORUM instance in the process while each audio stream remains inside its own track.

  • a.Zero Latency starts with EverClean to remove feedback and ringing while cleaning noise, bleed and room tail.
  • b.CALM assists sample by sample with a causal filter bank that detects narrow tonal build-up.
  • c.Studio 21 ms engages Mamba spectral control and more selective DNS for the hardest channels.
  • a.A single Clean Amount control scales the complete defeedback and source-cleaning response.
  • b.The direct path reports zero samples, making it suitable for monitoring-sensitive vocal channels.
  • c.Native mono, mono-to-stereo and stereo routing keeps microphone instances efficient and predictable.
  • a.SUBSTANCE fills the sub band without turning the whole master into a louder file.
  • b.IMPACT adds front-edge weight for kicks and bass drops in techno systems.
  • c.FINISH and GLOW are gain-conscious: polish, air and presence with beta-safe headroom.
  • a.WOW, WEIGHT, PUNCH, AIR and DEPTH move the mix psychoacoustically, not as blunt EQ boosts.
  • b.GUARD watches peak, loudness and translation so the bus stays safe in live routing.
  • c.Every module has bypass, the global Dry/Wet works instantly, and the engine stays at 0 samples.
  • a.16-line HyperFDN tank with geometry-modulated Householder feedback.
  • b.EDR tail view: the star dust is the actual wet decay by frequency.
  • c.Texture, smooth, glue and 4D motion keep the tail wide, soft and behind the source.
  • a.Up to 40.3% lower full-process CPU in the measured Apple Silicon comparison, with an optimized Windows x64 build.
  • b.Turbo extends Void Amount to 200% for deeper cleanup without adding latency.
  • c.Matter, Body, Consistency and four-band Matrix Control remain available in the measured 182-sample live path.
  • a.One Regeneration fader controls the complete reconstruction path.
  • b.Latency-aligned dry and regenerated signals preserve timing while the blend changes.
  • c.SOUL is designed after cleanup for post, playback and other latency-tolerant paths.
  • a.Nine tonal populations define a direction without replacing the source identity.
  • b.An 8192-sample analyzer informs a causal 30-band EQ while audio remains in the time domain.
  • c.Amount, Speed, Freeze and frequency bounds keep adaptation under direct control.
  • a.Seventy-five source profiles and twelve styles provide instrument-specific, musically contextual targets.
  • a.Pitch tracking follows fundamentals from 20 to 300 Hz and drives the generated sub and root layers.
  • b.ABYSS, ANCHOR, ROAR, FURY, GRIP and IMMERSION provide a complete low-end workflow from one view.
  • c.The advanced pages expose 82 parameters for tracking, synthesis, tone, dynamics and output safety.
  • a.Insert one instance per microphone, then choose the same Group and Link number on the channels that should cooperate.
  • b.EUCLID identifies the focused source using useful-voice evidence and temporal correlation; CLASSIC distributes gain proportionally from each channel's detected energy.
  • c.The model selection is global. Priority, Speed and Range remain local controls; Clean DNS is optional and never enters the shared bus.
  • b.H2-H5 harmonics remain visible in the RTA and controllable globally or across four bands.
  • c.Live reports 3 samples; Studio uses a 1920-sample phase-matched high-quality path at 48 kHz.
0 1 7 16 7 40 1 30 75 6 9 samples knob presets FDN macros % fader bands profiles macros parameters EverClean + CALM causal path focused source-cleanup control conservative beta factory set HyperFDN feedback lines livebus controls plus global mix lower live CPU measured against the previous release complete regeneration workflow causal adaptive EQ bands source-specific target profiles immediate low-end controls focused parameters per channel
21.3 0 0 5 0 3.8 96.7 0 3 0 0 ms samples samples modes samples ms ms samples samples samples samples CALM Mamba Studio path reported plugin latency zero-latency master insert Chamber · Plate · Hall · Nimbus · Infinity zero-latency direct livebus path 3.8 ms measured source path aligned regeneration path at 48 kHz zero reported latency reported latency in Live mode reported host latency reported host latency
100 % free after registration
/ 02How it works

Three stages. One signal path.

01 Listen

Choose the right engine Track the source and feedback Multiband sub analysis Psychoacoustic bus read Energy decay relief Optimized source cleanup Feed the cleaned source Long-term vocal observation Identify source and context Track the fundamental Publish local evidence

Use Zero Latency for monitoring: EverClean performs defeedback and source cleanup while CALM and DNS run causally with no lookahead. EverClean follows the incoming microphone, tracks emerging feedback and resonant ringing, and maintains an independent cleanup state for every plugin instance. MAGNETAR splits the master around the sub band and tracks low-end energy before adding anything, so the enhancement starts from the mix instead of a fixed tone. ATLAS follows crest, loudness, transient density and broad spectral balance so the macros react like a live engineer touching the bus. STARDUST reads the wet tail as a frequency-time field, so the visual fog follows the decay instead of decorating it. THE VOID V2 analyzes the incoming source and applies the original Everworks cleanup response through an optimized native engine. Place SOUL after THE VOID or another cleanup stage. Its reconstruction path receives the already-focused source instead of trying to clean it again. A periodic-Hann LTAS analyzer measures the stable vocal spectrum while the direct signal continues uninterrupted. Choose one of 75 source profiles and a musical context so the target is relevant to the instrument, register and production style. A guarded pitch detector follows the useful 20-300 Hz fundamental while the dual analyzer keeps input and output energy readable. Each instance measures input energy, useful voice, noise floor, confidence and an eight-bin temporal signature, then publishes only that compact metadata to its selected Group and Link.

02 Decide

Protect live or escalate One useful decision Coherence first Enhance under guard Dense, not metallic Extend cleanup depth Raise Regeneration Compare, then bound Match and shape harmonics Build and control weight Choose the group decision law

In Zero Latency, CALM tracks emerging narrow tones sample by sample. Switch to Studio 21 ms when difficult feedback or noise needs the full Mamba spectral model and higher-resolution DNS. Clean Amount scales the complete defeedback and source-cleaning response without exposing internal bands, timing or model controls. SUBSTANCE adds controlled sub support, IMPACT shapes punch, GLOW opens the top aura and FINISH glues the master with beta-safe headroom. WOW glues, WEIGHT anchors, PUNCH restores front edge, AIR opens presence, DEPTH widens perceived space, and GUARD keeps the bus inside live limits. Prime delay spacing, all-pass delay drift and a 4D-modulated Householder matrix build density while avoiding obvious ringing. Normal mode preserves the original 0-100% response. Turbo extends the range to 200% for sources that need stronger cleanup. The single fader moves through a latency-aligned blend in the processing domain, so Regeneration changes the reconstruction rather than acting as a simple output gain. D.N.A VOCALS compares the source with the selected tonal population and solves a restrained correction inside independent boost, cut and frequency limits. The adaptive EQ follows the selected target while H2-H5, motion, character and dynamics determine how harmonic energy is generated and distributed. Sub synthesis, root reinforcement and resonant tone are shaped through saturation and compression, with each stage available in the advanced pages. EUCLID compares useful-source evidence and rejects correlated spill. CLASSIC follows proportional Dugan-style sharing from channel energy. Changing the mode in one instance changes it across the entire linked QUORUM process.

03 Deliver

Keep the vocal usable Immediate output Translatable low end Direct live output Glue behind the source Same live timing, lower CPU Stop when it feels natural Causal adaptive output Choose Live or Studio Deliver a controlled low end Apply locally, share no audio

EverClean, CALM and DNS remain separately adjustable, so the channel can be cleaned and stabilized without treating every source as an extreme feedback case. The processed signal remains on a zero-reported-latency path with deterministic bypass and native mono or stereo routing. The output is gain-conscious for master bus use: more weight and presence, without pretending a fader lift is processing. The path is direct and zero-latency: no lookahead master limiter, no offline-style latency, just controllable live enhancement with Dry/Wet and instant bypass. Per-line damping, wet ducking and subtle saturation let the HyperFDN bloom musically without masking vocals or leads. Normal and Turbo both use the same measured 3.8 ms live path. Turbo adds no extra latency. Increase Regeneration only until body and stable detail return. SOUL is deliberately separate from the cleaner so restoration can be judged and bypassed on its own. A 30-band biquad bank moves at the selected speed in the time domain, so the adaptive EQ reports zero samples of latency. Live keeps monitoring immediate at 3 samples, while Studio switches to a 1920-sample phase-matched path for maximum rendering quality at 48 kHz. Mono control, output EQ, high-pass filtering, limiting and global mix keep the generated weight practical at zero reported samples. Speed smooths the local gain, Range limits attenuation and optional Clean DNS reduces local noise. Audio never leaves the plugin instance and the host sees zero samples of latency.

/ 03Features

Two modes, each with a clearly defined job. One-knob defeedback. No setup maze. Five core macros and a dual 8192-point RTA. Seven live controls for a bigger, safer bus. Six tools for a 4D musical space. V2: lower CPU with controllable Turbo depth. One fader for source regeneration. Six controls for adaptive vocal direction. Source-aware EQ and harmonics in one readable engine. Six macros above an 82-parameter low-end engine. Linked automixing without a central rack.

f.01

EverClean defeedback · Zero Latency Defeedback + Clean Amount SUBSTANCE WOW 16-line HyperFDN engine Optimized Everworks engine Regeneration Nine tonal targets 75 source profiles ABYSS · Sub Level One instance per channel

The recommended live workflow: EverClean suppresses feedback and ringing while cleaning noise, bleed and room haze with zero reported samples. One large control scales feedback suppression and source cleanup together, from untouched to fully engaged. Low extension and sub fill tuned for techno masters and large PA sub arrays. Psychoacoustic cohesion that makes a live mix feel more finished without needing master-bus latency. A 16-line feedback tank uses prime spacing, all-pass delay drift and dynamic Householder mixing to build density without metallic repeats. The original Everworks cleanup response runs locally in an optimized engine for macOS and Windows. One horizontal control introduces the complete regeneration result without exposing internal reconstruction parameters. Genome, Organic, Gravity, Neon, Pulse, Gloss, Depth, Focus and Air define the direction of travel. Detailed profiles cover voices, drums, basses, guitars, keys, orchestral instruments, electronic sources and full mixes. Adds a pitch-following sub layer with tracking offset, fallback tuning, gate depth and percussive response available underneath. Use QUORUM directly on each microphone track. Instances discover one another through the shared control bus, so no multichannel wrapper or central audio router is required.

f.02

CALM causal guard Zero reported latency IMPACT WEIGHT Spectral EDR fog Turbo · 100-200% Aligned reconstruction Amount Musical context ANCHOR · Root Tone EUCLID / CLASSIC global mode

A narrow-band causal filter bank follows tonal dominance and emerging ringing sample by sample, without FFT lookahead. The host receives zero samples of plugin delay, keeping monitoring and live vocal routing immediate. Transient weight for kick authority and drops, shaped to avoid broad master gain jumps. Low-gravity bus support for PA and stream feeds, tuned to feel solid before it feels louder. The analyzer draws the wet decay as a moving frequency cloud, making tail energy readable at a glance. At 200%, Turbo extends suppression depth while keeping the same reported latency and a protected output range. Dry and regenerated paths stay latency aligned, avoiding a timing jump as the fader moves. Blend from the natural source toward the continuously learned correction without replacing its identity. Twelve style contexts reshape the target for the role the source must play, rather than applying one generic curve everywhere. Reinforces the detected fundamental with controllable frequency, Q and gain instead of applying a broad fixed bass boost. EUCLID prioritises source focus and spill rejection. CLASSIC provides proportional Dugan-style gain sharing. Select either model once and every QUORUM instance follows it.

f.03

Dual-mode DNS Purpose-built routing FINISH PUNCH Early reflections Matter · Matrix · 182 samples Dedicated after cleanup Speed Visible H2-H5 engine ROAR · Resonance Spill-aware correlation

Five causal zones keep Zero Latency immediate; Studio uses spectral history and the Mamba mask for more selective noise suppression. Mono, mono-to-stereo and stereo layouts are handled explicitly, with one efficient pass for mono microphones. Master polish and glue inspired by high-end limiter workflows. Powerful by design; conservative by default. Transient focus for drums, playback and full-band buses, with safety compensation built around live gain staging. Deterministic stereo taps define depth in the first milliseconds while preserving the dry front edge. Matter and Body restoration, four-band Matrix tuning and Consistency stay inside the measured 3.8 ms live path. SOUL does not duplicate the cleaner. It is a focused second stage for post, playback and other latency-tolerant workflows. Slow, Medium and Fast determine how quickly the EQ follows a changing vocal performance. Second through fifth harmonics are measured in the RTA, shown with persistent trails and controlled independently. Shapes focused resonant character around the tracked low end, with drive, slope and mix controls for exact placement. Eight temporal energy bins let the group detect similar material arriving across microphones and avoid opening several channels for the same source.

f.04

Tonal / Low Lock GLOW AIR Damping and diffusion Freeze Motion and character FURY · Saturation Group A-D · Link 1-8

Focused protection for persistent ringing, podium microphones, lavaliers, headsets and low-frequency instability. Champagne air for presence and openness after the low-end work, without turning the plugin into a bright EQ. High-band clarity and expensive shine shaped for live systems, not a static top-end shelf. Per-line feedback absorption lets high frequencies die naturally before the body of the tail. Hold the learned correction when the vocal reaches the right balance, then release it when adaptation should resume. Transient or Sustain motion, Even or Odd character and linked or independent dynamics define how the harmonic layer behaves. Adds audible harmonic density with drive, bias, even/odd balance, shape, target range and selectable quality. Create independent automix pools inside one DAW or plugin server. Only instances with the same Group and Link number compete with one another.

f.05

CALM Mamba Studio 21 ms DRY / WET DEPTH Glue and ducking Frequency bounds Four bands and pitch follow GRIP · Compression Priority · Speed · Range

For difficult cases, Studio activates the full CALM Mamba network, high-resolution spectral suppression and the superior DNS path. Global parallel blend for instant level-conscious A/B decisions without leaving the core view. Perceived dimension and width enhancement for buses that need depth without reverb smear or added latency. Subtle saturation and envelope-aware wet ducking keep the reverb expensive, present and out of the way. Automatic or manual low and high edges plus independent maximum boost and cut keep correction focused. The advanced panel distributes harmonic processing across four bands and can track the fundamental of pitched material. Controls low-end dynamics with threshold, ratio, knee, attack, release, range and automatic makeup options. Bias important microphones, choose Fast, Natural or Smooth movement, and set the maximum attenuation without disturbing the shared detector logic.

f.06

Advanced vocal chain Dual RTA + A/B GUARD + Dry/Wet Five musical modes Mid and output Live, Studio and safety IMMERSION · Wet Mix Local Clean DNS

Expanse, Resonance, Clean EQ, Body, DeEss, compression and finishing remain available after the core cleanup decision. An 8192-point input/output analyzer, A/B snapshots, gain match, Delta and Live Lock make every move easy to verify. Live-safe macro with global blend and per-module bypass, so engineers can add only the lift the show can take. Chamber, Plate, Hall, Nimbus and Infinity cover tight vocals, lush plates and cinematic tails. Focus adaptation on the stereo centre and match the final level with output trim for honest A/B checks. Measured latency is reported to the host, with Delta, auto gain, mono safety and alias protection available for reliable A/B decisions. Blends the complete engine against the dry source while output EQ, mono control, filtering and limiting protect the result. A causal denoise stage works independently on each track after gain sharing. Its audio and state remain local and the path still reports zero samples.

/ 04Core modules

Eleven modules. One signal chain.

m.01EverClean zero-latency core
m.02CALM causal feedback guard
m.03CALM Mamba Studio engine
m.04Dual-mode DNS
m.05Auto Gain
m.06Tonal / Low Lock
m.07Expanse Gate
m.08Resonance control
m.09Clean Clarity EQ
m.10Body Low End
m.11DeEss Control
m.12Glue / Finisher
/ 04Core modules

MAGNETAR signal path. Five macros and live analysis.

m.01SUBSTANCE macro
m.02IMPACT transient weight
m.03FINISH master polish
m.04GLOW champagne air
m.05DRY / WET parallel blend
m.068192-point dual RTA
m.07A/B · Match · Delta · Live Lock
m.08AU · VST3 · Standalone
/ 04Core modules

LEVIATHAN signal path. Track, generate, shape and contain.

l.01Adaptive pitch tracker · 20-300 Hz
l.02ABYSS · pitch-following sub
l.03ANCHOR · root reinforcement
l.04ROAR · resonance contour
l.05FURY · harmonic saturation
l.06GRIP · low-end dynamics
l.07IMMERSION · global wet mix
l.08Output EQ · mono · HPF · limiter
l.098192 FFT · 128-bin dual RTA
l.10AU · VST3 · Standalone
/ 04Core modules

QUORUM signal path. Detect locally, choose globally, apply locally.

QUORUM 0.3 interface showing global EUCLID and CLASSIC automix selection
One global model selection · shared detector metadata · local audio only
q.01Local input · useful voice · noise-floor detector
q.02Eight-bin temporal channel signature
q.03Process-wide metadata bus · no shared audio
q.04Global EUCLID / CLASSIC decision model
q.05Relative-focus scoring · voice confidence
q.06Constant-power linked gain share
q.07Group A-D · Link 1-8 isolation
q.08Priority · Speed · Range controls
q.09Independent local causal Clean DNS
q.10AU · VST3 · zero samples reported
/ 04Core modules

STARDUST modules. Space that stays musical.

s.0116-Line HyperFDN Engine
s.024D Householder Matrix
s.03Tesseract Delay Drift
s.04High-Resolution EDR Fog
s.05Texture Diffusion
s.06Smooth Damping
s.07Glue Ducking
s.08Infinity Freeze
/ 04Core modules

ATLAS LiveBus modules. Psychoacoustic lift, live-safe control.

a.01WOW cohesion
a.02WEIGHT low gravity
a.03PUNCH transient lift
a.04AIR shine
a.05DEPTH dimension
a.06GUARD live safe
a.07Per-module bypass
a.08Global Dry/Wet
a.09A/B · Match · Delta
a.10TP · LUFS · SAFE meters
a.11Zero-latency direct path
a.12AU · VST3 · Standalone
/ 04Core modules

D.N.A VOCALS signal path. Deep analysis, causal audio.

d.018192-sample LTAS analyzer
d.02Tonal population matcher
d.0330-band response solver
d.04Causal biquad filter bank
d.05Target selector
d.06Amount blend
d.07Slow · Medium · Fast
d.08Freeze state
d.09Low / high frequency bounds
d.10Maximum boost / cut
d.11Mid focus
d.12Output trim
/ 04Core modules

D.N.A ULTIMATE signal path. Target intelligence plus harmonic design.

u.0175 source target profiles
u.0212 musical contexts
u.03Adaptive target EQ
u.04H2 · H3 · H4 · H5 engine
u.05Harmonic amount
u.06Transient / Sustain motion
u.07Even / Odd character
u.08Linked / Independent dynamics
u.09Four-band harmonic balance
u.10Pitch follow
u.11Delta · Auto Gain · Mono Safety
u.12Live / phase-matched Studio
/ 04Core modules

THE VOID V2. Original Everworks cleanup, with Turbo when needed.

THE VOID V2 with Matter, Turbo, Body, Consistency and Matrix Control
THE VOID V2 2.0.0 · optimized engine · Turbo extends Void Amount to 200%.
v.01Original Everworks cleanup engine
v.02Void Amount · normal 0-100%
v.03Turbo · extended range 100-200%
v.04Matter / Body · source restoration
v.05Consistency · target and speed
v.06Four-band Matrix · 3.8 ms live path
/ 04Core modules

SOUL signal path. One control, one defined job.

SOUL 0.1.0 regeneration plugin with one Regeneration fader
SOUL 0.1.0 · dedicated source regeneration after cleanup.
s.01Regeneration · complete amount
s.02Latency-aligned reconstruction
s.03Body and continuity recovery
s.0448 kHz processing path
s.05Mono and stereo operation
s.06Signed beta packages in preparation
/ ATLASLiveBus release

ATLAS LiveBus is ready for live tests.

Built from the zero-latency Magnetar base and redesigned for direct live buses: bigger perceived mix, safer output, instant bypass, Dry/Wet and clear TP/LUFS/SAFE feedback.

ATLAS LiveBus plugin screenshot with dual RTA
Release package
Latency0 samples
FormatsAU · VST3 · App
macOS13+ Universal
Windowsx64 VST3 · App
ValidatedAUVAL · Measure
Bus useMaster · Matrix · Stream
↳ Download for macOS ↳ Download for Windows
/ 04Open community

Not to sell, but to improve together.

⚡️

100% Free for Everyone

Register once and download the available Everworks plugins for free. No subscription, no watermark, no paid license. In return, we ask for real-world feedback from engineers and testers.

🛡️

Real-World Telemetry

Help us refine the Everworks engines. Share feedback on specific sources, microphones, rooms, stages and demanding setups.

🤝

Shape the Beta

Suggest features, vote on new DSP modules, and collaborate directly with the Everworks developers on future versions.

Open community board Join the Discord
/ 05Technical

Specs & compatibility.

Plugin formats

Formats
AU · VST3 · StandaloneAU · VST3
Binary
Universal (arm64 + x86_64)macOS Universal (arm64 + x86_64) · Windows x64Apple Silicon development build · release targets pendingmacOS ARM64 · macOS x86_64 · Windows x64
Package
macOS Universal · Windows x64Signed and notarized Universal DMG · Windows ZIPSigned macOS and Windows beta packages in preparationSigned and notarized Universal DMGSigned and notarized Universal DMGSigned + notarized macOS DMGs · Windows x64 installer
Modes
Zero Latency: EverClean + CALM causal · Studio 21 ms: CALM Mamba + spectral DNS Clean Amount · Active / Bypass Clean Master · Master Safe · Techno Weight · KS28 Sub Fill · Kick Authority · Air Glue · Festival Check WOW · WEIGHT · PUNCH · AIR · DEPTH · GUARD · Dry/Wet Chamber · Plate · Hall · Nimbus · Infinity Void Amount · Turbo 100-200% · Matter / Body · Consistency · Matrix Control Regeneration Genome · Organic · Gravity · Neon · Pulse · Gloss · Depth · Focus · Air 75 sources · 12 styles · H2-H5 · Live / Studio ABYSS · ANCHOR · ROAR · FURY · GRIP · IMMERSION EUCLID / CLASSIC global mode · Automix · Group A-D · Link 1-8 · Priority · Speed · Range · Clean DNS

Audio engine

Sample rates
44.1 / 48 / 88.2 / 96 kHz 44.1 / 48 / 88.2 / 96 kHz 44.1 / 48 / 88.2 / 96 kHz 44.1 / 48 / 88.2 / 96 kHz 44.1 / 48 / 88.2 / 96 / 176.4 / 192 kHz 44.1 / 48 / 88.2 / 96 kHz 48 kHz 44.1 / 48 / 88.2 / 96 / 176.4 / 192 kHz 44.1 / 48 / 88.2 / 96 / 176.4 / 192 kHz 44.1 / 48 / 88.2 / 96 kHz 44.1 / 48 / 88.2 / 96 kHz
Bit depth
32-bit float internal
Channels
Mono · Stereo
Latency
Zero Latency: 0 samples · Studio: 21.3 ms @ 48kHz 0 samples reported · direct cleanup path 0 samples · 32-bit float internal · 44.1-96 kHz 0 samples · direct livebus path · 32-bit float internal Zero dry latency · 16-line HyperFDN · 44.1-192 kHz tested 182 samples / 3.8 ms @ 48 kHz · unchanged with Turbo 4641 samples / 96.7 ms @ 48 kHz · aligned regeneration path 0 samples reported · causal 30-band EQ · 44.1-192 kHz tested Live: 3 samples / 0.06 ms · Studio: 1920 samples / 40 ms @ 48 kHz 0 samples reported · direct adaptive low-end path 0 samples reported · linked metadata updates on the next host callback

System

macOS
13 Ventura or newer 13 Ventura or newer · Universal 13 Ventura or newer 13 Ventura or newer 13 Ventura or newer 13+ Intel · 14+ Apple Silicon beta macOS beta package in preparation 13 Ventura or newer · Universal 13 Ventura or newer · Universal 13 Ventura or newer · Universal 13 Ventura or newer · separate Apple Silicon and Intel packages
Memory
Zero allocation (audio thread)
Precision
float32 deterministic
CPU
Optimized for Apple Silicon · Intel supported Apple Silicon · Intel · Windows x64 Optimized for Apple Silicon · Intel supported Optimized for Apple Silicon · Intel supported Optimized for Apple Silicon · Intel supported Optimized native processing · up to 40.3% lower measured on M3 Max · optimized Windows x64 Apple Silicon development build Apple Silicon + Intel Apple Silicon + Intel Apple Silicon · Intel · Windows x64 Apple Silicon · Intel · Windows x64

Release package

macOS AU
Logic Pro compatible
macOS VST3
Ableton · Reaper · Cubase compatible hosts
Windows VST3
VST3 compatible hosts
Standalone
Included for quick local testingNot included · designed for one plugin instance per track
/ 06CALM instructions

A complete guide for beta testers.

Visual user guide · EN

Every CALM EverClean control, explained with screenshots.

This guide is built for testers and live engineers. It explains the top bar, CALM, EverClean, Expanse, Resonance, Clean EQ, DNS, Body, DeEss, Vocal Finisher, Glue, presets and output trim, with one screenshot per module.

  • Module-by-module screenshots
  • What every button, knob, fader and selector does
  • Studio 21 ms vs Zero Latency workflow
  • Soundcheck recipes and troubleshooting
↳ Open visual guide Markdown copy
CALM EverClean full interface with the redesigned channel strip open
/ 06THE VOID V2 instructions

V2 optimized processing, Turbo, Matter, Consistency and Matrix.

THE VOID V2 2.0.0 · current controls

The original Everworks response with a faster engine and new Turbo range.

V2 preserves the original Everworks model and live timing in an optimized engine. Turbo changes Void Amount to 100-200% for stronger cleanup without changing reported latency.

  • Normal Void Amount from 0-100%
  • Turbo Void Amount from 100-200%
  • Matter enable and separate Body amount
  • Consistency Target and Speed
  • Matrix 1-4 amount, trim, bypass, solo and crossovers
↳ Open V2 visual guide
THE VOID V2 interface with Turbo enabled
/ 07CALM documentsEverClean documentsTHE VOID V2 documentsSOUL documentsD.N.A VOCALS documentsD.N.A ULTIMATE documentsMAGNETAR documentsATLAS documentsSTARDUST documentsLEVIATHAN documentsQUORUM documents

User guides and workflows.

User guide · EN CALM EverClean visual instructions

Module-by-module screenshots and clear explanations for every visible control.

 Quick start · EN / ES EverClean one-control workflow

Signal path, Clean Amount, bypass, routing, latency and supported packages.

 THE VOID V2 2.0.0 · EN THE VOID V2 architecture and controls

Original Everworks cleanup, Turbo 100-200%, Matter, Body, Consistency, Matrix Control and measured latency.

 SOUL 0.1.0 · EN / ES SOUL regeneration overview

One-control workflow, aligned reconstruction, 48 kHz operation and current beta packaging status.

 Architecture · EN / ES D.N.A VOCALS adaptive EQ overview

LTAS observation, nine tonal populations, 30-band causal solver, controls, formats and latency.

 Architecture · EN / ES D.N.A ULTIMATE engine overview

Source profiles, styles, target matching, visible H2-H5 harmonics, Live and Studio signal paths.

 Architecture · EN / ES LEVIATHAN adaptive low-end overview

Pitch tracking, six macros, advanced Track/Sub/Tone/Dynamics/Output pages, analyzer, latency and release formats.

 QUORUM 0.3.2 · EN / ES Distributed automix architecture

One instance per channel, global EUCLID/CLASSIC selection, spill correlation, proportional gain sharing, Group/Link routing, local Clean DNS and release packages.
/ 08FAQ

Things engineers ask.

Does it colour the source?

Both paths are dynamic rather than fixed EQ. Start with EverClean in Zero Latency, add only the CALM protection you need, and reserve Studio Mamba for difficult channels. EverClean is a cleanup processor, not a tone preset. Start low, level-match the result and stop when the source is clean enough. MAGNETAR is designed to keep the master musical. SUBSTANCE and IMPACT should feel bigger before they feel louder; FINISH is intentionally powerful and best used conservatively in the beta. ATLAS is intentionally audible when you push it, but it is built for live translation: the macros add perceived finish while GUARD, Dry/Wet and per-module bypass keep the bus under control. It is designed to add space without stealing the dry source. Damping, ducking and glue keep the tail behind the vocal or lead. THE VOID V2 is built as a cleaner, not a tone effect. Push Void Amount until the source tightens; enable Matter and add Body only when the cleaned signal needs reconstruction. SOUL is a regeneration processor rather than an EQ preset. Level-match the result and stop when body and stable detail return naturally. D.N.A VOCALS applies a bounded correction and keeps the source as its starting point. Amount determines how far the vocal moves toward the selected tonal population. D.N.A ULTIMATE starts from a source-specific target, then lets you add and shape H2-H5 harmonics. Amount and harmonic controls remain independent, with Delta and auto gain for honest checks. LEVIATHAN is deliberately audible, but its generated sub, root, resonance, saturation and compression are independent. Start with the six macros low and verify with IMMERSION at matched level. Automix changes level by design, not tone. EUCLID is the spill-aware focused model; CLASSIC is proportional Dugan-style gain sharing. With Clean DNS at zero, neither mode adds local denoise.

Can I run it on in-ear monitor mixes?

Yes. Use EverClean with CALM's causal guard in Zero Latency for monitoring. Use Studio 21 ms only when the route can accept latency and needs the Mamba engine's extra selectivity. Yes. EverClean reports zero samples of plugin latency and is designed for monitoring-sensitive microphone paths. Yes. Use conservative settings first and A/B against bypass at matched level. Yes. Use conservative settings first and A/B against bypass at matched level. Yes. Use conservative settings first and A/B against bypass at matched level. Yes. THE VOID V2 has measured 3.8 ms in real testing and is intended for monitoring-sensitive routing. Not on a latency-sensitive monitor path. SOUL reports 4641 samples at 48 kHz, approximately 96.7 ms, and is intended for post, playback or latency-tolerant routing. Yes. The analyzer informs a causal time-domain EQ, so D.N.A VOCALS reports zero samples of latency. Yes. Live mode reports only 3 samples at 48 kHz. Studio mode reports 1920 samples and is intended for phase-matched high-quality rendering. Yes. LEVIATHAN reports zero samples of host latency. Its low-end generation is designed for tracking, live playback, buses and studio inserts. Yes. QUORUM reports zero samples of latency. For monitor mixes, use a dedicated Group/Link and a conservative Range, or disable Automix on channels that must remain fixed.

Will it work on broadcast and corporate gigs?

Yes. CALM is designed for corporate, broadcast and live workflows where lectern mics, lavs and stage microphones need stability without obvious processing. Yes. Its one-control workflow is intended for fast deployment on lecterns, headsets, lavaliers and vocal microphones. Yes. The beta is intended for real production testing, and we want host/source feedback. Yes. The beta is intended for real production testing, and we want host/source feedback. Yes. The beta is intended for real production testing, and we want host/source feedback. Yes. It is especially relevant for lecterns, lavaliers, headsets and vocal mics in reflective or noisy rooms. Yes, after cleanup and where its reported latency is acceptable. It is designed as a separate reconstruction decision, not as the first live insert. Yes. It is designed for vocal sources in live, broadcast and studio workflows; Freeze can hold a learned correction when the balance is right. Yes. Use Live for tracking, broadcast and performance; use Studio when latency is secondary to maximum rendering quality. Yes. Use conservative ABYSS and ROAR values on full-range systems, then open the advanced output page for mono control, filtering and limiting. Yes. Lecterns, lavaliers, headsets, panels and broadcast microphone groups are its primary workflow: one instance per channel, all assigned to the same Group and Link.

How is it licensed?

Yes. Register free to unlock the download. No payment, no subscription, no watermark.

How can I help improve it?

Run it on real shows, tell us what works, report edge cases and suggest new modules. That is the exchange: free access for useful real-world feedback.

// Free with registration · Full features

Run the next show with CALM on every channel.

SOUL beta packages are being prepared.

Create a free account to unlock the download. No payment, no subscription, no watermark.

↳ Download for macOS ↳ Download for macOS Intel ↳ Download for Windows ↳ Windows portable VST3 (.zip)
QUORUM 0.3.2 Plugin Nap resilience keeps linked instances and global mode changes alive while hosts suspend silent audio. macOS ARM64 and Intel are signed, notarized and stapled; Windows x64 is tested but currently unsigned.
THE VOID V2 2.0.0 Original Everworks cleanup, Turbo 100-200% and 3.8 ms live latency. macOS packages are signed, notarized and stapled. Windows x64 is optimized and currently unsigned.
EverClean 1.0.1 Windows CPU update Static oneMKL backend measured 24.2% lower mono CPU and 21.7% lower stereo CPU in the same-system A/B. No external MKL DLL is required.
Signing pending The plugin page is live; official macOS and Windows beta installers will be added after signing and validation.

Everworks Suite · AU · VST3 · Standalone · available builds vary by product · download unlocks after free registrationQUORUM 0.3.2 · AU + VST3 on macOS · VST3 on Windows · download unlocks after free registration