VESTIGE is a stereo algorithmic reverb whose echo tap sits between the pre-delay and the diffusers, so every repeat re-enters the room as a fresh strike rather than arriving as a delay bolted on in front of it.
Four engines: DENSE, MOD, PLATE and VOID. A SOURCE control that places the player near or far, DRIVE into the tank's own saturation, a tempo-locked echo with its own colour, and a tank you can freeze. No convolution and no impulse library: every tail is generated. Zero latency, at every setting.
An algorithmic reverb builds a room out of delay lines that feed back into each other. Every pass round the loop loses a little level, and how fast it loses it is the decay time. VESTIGE's tank is four delay lines of mutually incommensurate length, mixed by a Householder matrix, each carrying its own nested allpasses, so echo density grows with every circulation instead of arriving flat.
Decay runs in three independent bands. The mids follow DECAY on Jot's per-line law. Below 480 Hz the decay is scaled by TONE, from twice as long as the mids to half as long. The highs are rolled off by DAMPING inside the loop, so they can only ever die faster than the mids, never slower.
The difference is where the echo lives. Its tap sits after the pre-delay and before the four input diffusers, so each repeat is diffused and fed into the tank as a new strike. The echo loop closes on its own buffer and never through the tank: feeding a reverb back into its own input oscillates long before a feedback control reaches its top.
Every recirculation write is soft-saturated and rounded to a 26-bit fixed-point word. The rounding is deliberate. It is the arithmetic character of the tank, and it is what flushes the tail to true zero, so silence decays to silence rather than to a denormal floor.
The input is summed to mono, DC-blocked and band-limited at 16.5 kHz. DRIVE, when it is up, pushes it into the tank's own saturation curve.
PRE-DELAY holds it back up to 250 ms. SOURCE adds or takes away the time of flight from the player to the room.
The echo tap writes repeats that will each enter the room as a new strike, every TIME, darkened by COLOR.
Four allpasses in series smear each strike before the tank. DIFFUSION sets how hard.
Four lines and their nested allpasses, mixed by a Householder matrix, decaying in three bands.
The output is read from distinct points in the tank, so left and right are incoherent. Tail correlation measures −0.022.
This is VESTIGE's decay arithmetic running in your browser: the same four line lengths, the same nested allpasses, the same per-line law, damping filter and echo pole the plug-in computes when you turn a knob. It draws the energy left in three bands after a single strike, in decibels, against time. The vertical ticks along the top are the echo's repeats; each one strikes the room again, which is why the curves step back up.
ENGINE is structural. Switching it fades the output out, rebuilds the tank and fades back in. Each card draws the low, mid and high energy of the same strike through that engine at VESTIGE's default settings, from the decay model above.
No delay-line modulation at all. The tank carries 27% more line length and one extra nested allpass per branch instead: density bought with memory rather than movement, so pitch-stable material stays pitch-stable. SPIN and WANDER go inert, and the panel says so.
Independent band-limited random-step walks on every line sweep the mode picket, so a sparse network time-averages into a dense one. SPIN is the step rate and WANDER the depth. The pitch movement is capped at 3 cents a pass, so a long tail never drifts out of tune with the dry signal.
A 56-stage dispersive allpass cascade ahead of the tank. High frequencies arrive first and the initial transient is smeared before density builds. DAMPING sets the metal, not the air. The cascade's colouration above 2 kHz is its own signature and does not move for any tank setting.
Nothing arrives early. Six long allpasses ahead of the diffusers delay the energy without colouring it, so the tail rises for a few hundred milliseconds before it falls. It runs DENSE's longer tank, closes the input at 4.5 kHz, puts the in-loop damping an octave below the knob and weights the output to the sides, so the centre is left empty. WANDER moves its level; its pitch movement is held to MOD's ceiling.
What makes VOID different is measured, not asserted. Against MOD at the same settings:
Where VOID's energy peaks, against 40 ms for MOD.
Left/right correlation of the tail, against −0.01 for MOD.
Centroid of the wet signal under sustained noise, against 3.5 kHz for MOD.
VOID's wet loses about 2.5 dB more than MOD's when folded to mono. That is the price of an empty centre, and it is by design.
Press SHOW IN THE LAB on any card and the decay lab above sweeps that control through its range. Size, decay and damping changes glide; a preset recall or an automation step on TIME crossfades instead, so the echo can never chirp. Every neutral position is exactly neutral: FEEDBACK 0, SOURCE 50 and DRIVE 0 are bit-identical to the signal path before those controls existed.
DECAY is the hero: 0.20 to 20 s for the mid band, skewed so the short end is not crowded into the first few degrees. SIZE runs 8 to 100 m and scales every delay in the tank, the early reflections and the time of flight with it. Dragging SIZE glides the lines rather than jumping them.
Places the player in the room, near (0) to far (100). Near brings the room in sooner and puts eight discrete stereo early reflections in front of the tail: the first 60 ms measure +4.9 dB. Far brings it in later, darkens the input with air down to about 2.9 kHz, smears the onset and brings the tail back up to level: the 8 kHz octave falls 13.9 dB while the steady level holds within 0.5 dB. The time of flight follows SIZE: ±25 ms at SIZE 34, ±73 ms at SIZE 100.
Pushes the input up to +18 dB into the same saturation curve the tank uses on its own recirculation, level-matched at −12 dBFS, so the room is struck harder rather than made louder. At 0 the stage is not run at all. It is not oversampled.
Two controls for the weight of the tail. TONE sets the low band's decay against the mids below a fixed 480 Hz: at 0 the bottom rings twice as long, at 100 half as long, and 50 is neutral. DAMPING, 1 to 20 kHz, is the cutoff of the filter inside the loop, so it shortens the top on every pass.
The echo row. TIME runs 20 ms to 2 s, and SYNC snaps it to the nearest of eleven divisions at the host tempo. FEEDBACK runs 0 to 100%; at 0 the branch is skipped, not run quietly. COLOR voices the repeats ±3 octaves around the room's own darkening, independent of the tank, and 50 is neutral.
PRE-DELAY holds the room back 0 to 250 ms. DIFFUSION sets how hard the four input allpasses smear each strike before the tank: low, the room answers with a handful of discrete reflections; high, they blur into one onset.
The modulation, on MOD and PLATE. SPIN is the rate of the random-step walks, 0 to 3 Hz; WANDER is their depth. The walks are what keep a four-line tank from ringing, so they are load-bearing rather than decorative. On DENSE they go inert; on VOID WANDER moves level instead.
An equal-power blend. At MIX 0 the dry signal is left alone: the residual is −300 dB, the measurement floor. BYPASS is a click-free crossfade with the tank left running underneath, nulling at −151.2 dB.
Freezes what the tank is already holding and stops new input going in, so you can play over the result. The echo keeps running underneath, so a tempo-locked delay keeps its place through the freeze. HOLD is a performance switch and is not stored in presets.
A delay placed in front of a room arrives as a delay. VESTIGE puts it inside the room: between the pre-delay and the diffusers. Each repeat is diffused and enters the tank as a new strike, so a dotted-eighth echo becomes a dotted-eighth room, each one building its own tail on top of the last.
Dragging TIME glides the tap the way tape does. A preset recall or an automation step crossfades instead, so the tap can never chirp. A 400 ms tap arrives at 406.8 ms, measured. With SYNC on, TIME keeps its milliseconds and only the readout changes, so automation stays continuous and a preset reloads the same with or without a host tempo. The divisions run 1/32 to 1/2, with triplets and dotted values, and the snap works in the log domain so it feels the same across the whole sweep.
Damping is applied per unit of time, not per repeat. The tank recirculates roughly every 52 ms; the echo passes once per TIME. Sharing one coefficient would darken the repeats about fourteen times more slowly than the room and grow a bright tail that outlives it. So the echo's pole is the tank's cutoff scaled by how much longer its own loop is, and at COLOR 50 the repeats lose highs at the same rate per second as the room.
COLOR moves that pole up to three octaves either way: darker than the room for a tape throw, brighter than it for a lit-up one. Between its extremes it spans 56.0 dB above 6 kHz. At 50 it is neutral and bit-exact.
Every factory preset ships with FEEDBACK at 0, so recalling one never starts a regenerating delay you did not ask for.
Press HOLD and the line gains cross to just under unity, the in-loop filters and the write-back saturator blend out, and new input stops going in. What the tank was holding stays there, and you play over it. The display draws the change: the decay envelope runs flat, the rays from the source dim because nothing new is arriving, the field reads HELD, and the RT60 readout says HELD instead of a number.
Lift it and the tank decays from where it was, with the echo still in time.
A held tank is a multi-minute sustain, not a mathematical freeze, and the figures say how long. Over half a minute frozen the low band falls 0.62 dB and the high band 13.30 dB, against −202 dB for the same material with HOLD off. The fundamentals hold for tens of minutes; the top settles over a couple of them.
The field in the middle of the panel is the room, drawn. Time runs away from you to the right, frequency band runs across to the left, and energy stands up out of the plane. The source strike hangs above the field as a single gold hub, and every line that falls from it is part of what VESTIGE is doing to your signal right now.
It is a still frame, 186 filaments from a fixed seed, re-drawn only when a control that shapes it moves. It never animates on its own. The one live part is height: the filaments rise and fall with the real wet level, band by band, thirty times a second. Each element answers to specific controls, so turning a knob moves the part of the picture it belongs to and nothing else.
Under the field sit three readouts. ECHO is the density of reflections per second, MODES the modal density per 100 Hz, and RT60 the decay of the composite tail, so a long regenerating echo is not hidden behind a short dialled decay.
MODES and RT60 are read from the tank's actual configuration. ECHO is an analytic estimate from the same tables, not a measurement. SPIN, WANDER, DRIVE, COLOR and MIX are not drawn.
The trial is the whole plug-in for fourteen days, and the header counts down the days left. After that, a computer that has not been activated goes silent until a licence is activated. In GarageBand and other sandboxed hosts it keeps playing.
The same installer everyone buys, running unlicensed for fourteen days. A licence unlocks it in place, no reinstall.
Activation needs an internet connection once. After that VESTIGE runs offline, and the licence does not expire. One licence covers three computers, one activation per computer however many user accounts it has. DEACTIVATE THIS COMPUTER frees a seat, so you can move VESTIGE to a new machine without writing to us.
Once activated, VESTIGE is never silenced by a licence check. After a hardware change or a system reinstall it keeps playing, and the header asks you to reactivate.
Zero latency on every engine, a dry path that MIX 0 leaves alone, and decay that runs in three bands, so the low end and the top can be set against the mids rather than with them.
A tempo-locked echo that strikes the room on the grid, with its own colour, and HOLD to freeze a chord and play over it while the echo keeps time underneath.
VOID's late, wide swell, FEEDBACK past the point where repeats stop dying, DRIVE into the tank's own curve, and twenty-second decays that HOLD can sustain for minutes.
SOURCE moves a voice or an instrument from close to distant without re-patching, and SIZE scales the whole space, reflections and time of flight together.
| TYPE | Four-engine stereo algorithmic reverb with an internal regenerating echo (insert) |
|---|---|
| STRUCTURE | 4-line feedback delay network, Householder mixing, 4 series input diffusers, 2–3 nested allpasses per line, 26-bit internal recirculation word. No convolution, no oversampling |
| ENGINES | DENSE · MOD · PLATE · VOID, structural, with a fade-rebuild-fade swap |
| DECAY | 0.20–20.0 s in three bands: mid by DECAY; low below 480 Hz ×2.0–×0.5 by TONE; high by DAMPING (1–20 kHz) inside the loop |
| ROOM | SIZE 8–100 m · PRE-DELAY 0–250 ms · DIFFUSION 0–100 % · SPIN 0–3 Hz · WANDER 0–100 % |
| SOURCE | Near to far; ±25 ms of flight at SIZE 34, ±73 ms at SIZE 100; eight stereo early reflections near, air to 2.9 kHz far |
| DRIVE | 0 to +18 dB into the tank's saturation curve, level-matched at −12 dBFS |
| ECHO | TIME 20 ms–2 s · FEEDBACK 0–100 % · COLOR ±3 octaves · SYNC to 11 divisions (1/32 to 1/2, triplets and dotted) |
| HOLD | Freezes the tank; low band within 0.62 dB over 30 s |
| OUTPUT | MIX equal-power 0–100 % · click-free BYPASS |
| CHANNELS | Stereo in, stereo out |
| LATENCY | Zero, in every configuration and at every setting |
| SAMPLE RATES | 44.1–192 kHz, wet level matched across them |
| CPU | 0.0027 of realtime (DENSE, MOD) to 0.0059 (PLATE) at 48 kHz / 512, one instance |
| DISPLAY | Impulse scatter, 186 filaments, live wet level at 30 Hz · ECHO, MODES, RT60 readouts · window 1024 × 710, resizes 75–150 %, size remembered |
| PRESETS | 13 factory (VAULT DOOR, STONE CHAMBER, GLASS HALL, TAPE PLATE, NIGHT AIR, BRICK ROOM, LONG NAVE, VELVET ROOM, STILL WATER, BRASS PLATE, IRON SHEET, SOFT FOCUS, GHOST CHOIR), also as host programs; user presets with save and save as, shared across your hosts |
| LICENCE | 14-day full trial · one licence, three computers · activate once online, then runs offline · perpetual |
| FORMATS | VST3 · Audio Unit · AAX (Pro Tools) |
| SYSTEM | macOS 11 or later · Apple Silicon and Intel (universal) · Developer ID signed, notarized Windows (VST3): coming soon |
Every tail VESTIGE makes is generated by the tank as you play. There is no impulse-response library and nothing is sampled, so every control moves the room itself rather than choosing between recordings of rooms.
The CPU figures were measured on a Mac before SOURCE and DRIVE were added; VOID costs about 18% more than MOD.
VESTIGE is calibrated and available for immediate deployment. All four engines, the echo, HOLD and the full factory bank, nothing withheld.
Launch sale. Use code ALCHEMY50 at checkout. The trial is the same installer, unlicensed for fourteen days; a licence unlocks it in place, no reinstall.