Welcome to modouVerb

modouVerb is a space-based, high-density reverb from modou dsp.

Most reverbs give you a set of knobs such as size, decay, diffusion and mix, and leave you to imagine the room. modouVerb works the other way round. You choose a space that exists as real 3D geometry. You place a sound source and a microphone inside it, and you decide how far apart they stand. The engine then renders what that microphone hears at that spot, facing that way, in that room.

modouVerb comes in three editions built on one engine:

• modouVerb Standard is the full mono and stereo reverb.
• modouVerb Immersive is the full reverb for 7.1.4 to 9.1.6 speaker beds.
• modouVerb Lite is a simple version for musicians who just want it to sound right.

This manual describes modouVerb v1.1.4.

A reverb is heard at one particular spot in a particular room. When you stand closer to a singer, three things change at once:

• the direct sound gets louder and brighter;
• the first reflections arrive later relative to it, and in a different order;
• the reverberant tail stays roughly as loud as before.

Turn your head, and the reflections swap ears. Move the singer toward a side wall, and that wall answers first and loudest.

Classic reverbs imitate the result of these relationships with independent controls such as mix, pre-delay, size and width. You can set those controls into combinations that no real room could produce. Sometimes that is exactly what you want. But it is also why a reverb so often sounds "pasted on" instead of around the source.

modouVerb keeps the relationships and lets you control the causes:

Where is the source? Where is the microphone? Which way does it point? What room are they in?

Everything you hear follows from those answers: the level and colour of the direct sound, the timing and direction of every early reflection, the balance of early to late energy, and the image across the speakers.

Requirements: macOS 10.13 or later on Intel, macOS 11 or later on Apple Silicon, and a host that loads Audio Unit (AU), VST3 or AAX plugins. Each edition also ships as a standalone application. Run ./build_installer.sh for a signed, notarized installer package.

In the host, the plugins appear under the manufacturer modou dsp. The three editions are separate plugins with separate plugin codes. You can install them side by side and use them in the same session.

User presets are saved as .mverbpreset files in ~/Library/Audio/Presets/modou dsp/modou VERB/. Standard and Immersive share this folder.

If a host still shows an older version after an update, rescan the plugins. On macOS you can also log out and back in so the Audio Unit cache is rebuilt.

Licensing

Each edition needs a licence from modou Central before it processes audio; an unlicensed instance stays silent (a mute DSP path, not a broken one). Open the gear icon in the plugin's header (or the modou Central app) to enter a licence key and email, or to request a trial. A licence is checked against your machine's hardware ID and can be moved between machines by deactivating it first. The three editions use separate SKUs (modou0010 Standard, modou0011 Immersive, modou0012 Lite), so each is licensed on its own.

When a sound plays in a room, the microphone (or your ear) receives three kinds of energy, one after another:

1 · Direct sound

The sound that travels in a straight line from the source to the microphone. Its level falls by 6 dB for every doubling of distance (the inverse-square law), and air absorbs its highest frequencies as the path gets longer. The direct sound tells the brain where the source is.

2 · Early reflections

Sound that bounced once, twice or three times off the walls, floor and ceiling before it reached the microphone. Each reflection has a precise arrival time, a direction and a colour, set by the path length and by the surfaces it touched. The gap between the direct sound and the first reflection is the initial time delay gap (ITDG). The ear reads the pattern of early reflections as the size and shape of the room and as the distance to the source.

3 · Late reverberant field

After many bounces, the reflections become so dense that they merge into a smooth, decaying wash with no audible direction. Its decay time T60 (the time to fall by 60 dB) depends on the room's volume and on how absorbent its surfaces are, separately at low, mid and high frequencies. In a given room the late field is nearly equally loud at every seat.

The critical distance

Because the direct sound falls with distance but the late field does not, there is one distance at which they are equally loud: the critical distance r_c. Closer than r_c, you hear mostly the source. Farther away, you hear mostly the room. In a small studio r_c is about a metre. In a large, reverberant hall it is several metres. This single fact is why "wetness" in a real room is a matter of where you stand, not of a mix control.

Almost every algorithmic reverb descends from the same family of ideas. Schroeder's comb-and-allpass networks came in the early 1960s. Moorer added early-reflection taps and in-loop filtering. From the 1990s on, the feedback delay network (FDN) became the standard: a set of delay lines whose outputs are mixed by a matrix and fed back. These are excellent tools. modouVerb itself uses a high-density FDN for its late field. But when the whole reverb is built only from delays, several limits follow.

The room is a statistic, not a place

The delay lengths are chosen to be mutually prime and "roughly room-sized". A SIZE control usually scales all of them together. Nothing in the algorithm knows where the walls are. So nothing can tell you what happens when the singer moves to the left, or when the microphone moves back.

Early reflections are a fixed tap pattern

The early part is usually a tapped delay line: a hand-made list of times and gains, sometimes one per "room type". The pattern is the same for every source and every listening position. In a real room, moving the listener changes every reflection's time, level and direction. In a tap pattern, only the global PRE-DELAY moves.

DRY/WET is not distance

A mix knob crossfades between the dry signal and a finished reverb. In a real room, moving away from the source changes several things at once:

• the direct sound drops and loses high frequencies;
• the reflections arrive closer to the direct sound in time;
• the balance of early to late energy shifts;
• the late field stays almost constant.

A mix knob does only the first of these, and without the colour change. That is why turning up the mix often sounds like "more reverb on top" rather than "further away".

Ringing, flutter and metallic colour

A network with too few echoes per second rings at its own delay lengths. Designers hide this with modulation, which smears the pitch, and with heavy input diffusion, which smears the transients. The tail then sounds smooth but generic.

Stereo is a trick, and multichannel is an upmix

Stereo width usually comes from taking two decorrelated outputs of the same network. There is no microphone that can be turned, no polar pattern, and no image. Most "surround" reverbs create extra outputs by further decorrelation or by upmixing stereo. The reflections do not come from where the walls are.

A delay-based reverb answers "how long and how dense?" very well. It cannot answer "from where, to where, and in what room?"

Convolution reverbs take the opposite approach. A real room is measured once: a loudspeaker plays a sweep, a microphone records it, and the result is the room's impulse response (IR). Convolving your audio with that IR places it in the measured room with great realism. For that one seat, convolution is the most faithful method there is. The limits come from the word one.

One source position, one mic position, frozen

An IR is valid for exactly one loudspeaker position and one microphone position. You cannot move the singer closer, move the mic back, raise it, or place the drums at the side of the stage. Libraries try to cover this with a few alternative positions (near, mid, far). Switching between them jumps, it does not move.

The measurement equipment is baked in

The IR contains more than the room:

• the frequency response and directivity of the loudspeaker used;
• the microphone model, its polar pattern and the direction it faced;
• the stereo technique (AB, XY, ORTF, Decca tree);
• the measurement noise floor and any truncation of the tail.

None of these can be changed afterwards. You cannot turn a cardioid into a figure-8, or rotate the mic by 90°.

Crossfading two IRs is not an intermediate seat

Blending the "near" and "far" IRs gives you both sets of reflections at once, at their own times: a double image with comb filtering. It is not the reflection pattern of a seat in between. Distance cannot be automated smoothly with fixed IRs.

Editing the IR is not physics

Shortening or lengthening the decay by reshaping the IR's envelope also reshapes the early reflections and bends the natural frequency-dependent decay. Pre-delay made by trimming or padding the IR moves the direct sound together with the room.

Multichannel multiplies the problem

A stereo-in, stereo-out ("true stereo") reverb already needs four IRs. For a real 7.1.4 input and output you need one IR from every input speaker to every output microphone: 12 × 12 = 144 measured responses, for one room and one arrangement. They are rarely measured, rarely consistent, and very expensive to convolve. So most immersive convolution reverbs place one or two sources in the room and upmix the rest.

An impulse response is a photograph of a room. modouVerb is the room.

modouVerb models the real physical conditions of a space — its geometry, its materials, where the source and the microphone actually stand in it — and generates the reverb from that model, rather than approximating a room with a chain of delay taps or freezing one seat into a fixed impulse response. It is a reverb built on real acoustic conditions carried as far into the virtual as they will go, which is what puts it past the fixed limits of an ordinary algorithm and of an IR. Concretely, it combines the physical parts that matter for where with a high-density network that is best at how long. Room-acoustics prediction software uses the same hybrid.

Because the reflections are computed rather than recorded, you can change what an IR cannot: move the source, move and turn the microphone, change its pattern, lengthen the decay without touching the early field, and render from a 2-channel to a 16-channel speaker layout from the same model.

Because the late field is a tuned high-density network rather than a recording, it is noise-free, can be as long as you like, reacts smoothly to every control, and is cheap: about 1–2.5 % of one CPU core for a stereo instance.

The last row is honest. If you need that exact hall, convolution is the right tool. If you need a room you can work in, where you place things, move them and automate them, modouVerb is.

Most musicians do not want to think about volumes, materials and polar patterns. They want the vocal "a bit further back, in a nice hall". Lite is built for exactly that. It has three controls, plus BYPASS:

• Place: ROOM, HALL, CHURCH or PLATE.
• Size: SMALL, MEDIUM or LARGE — scales that place's room up or down, the same way DECAY does in Standard and Immersive.
• DISTANCE: a big dial from CLOSE to FAR, showing metres (centimetres on the Plate).

Everything else follows the place and size. It runs the full modouVerb engine with a microphone chosen automatically: a cardioid, as an XY pair on a stereo output, in the space's designed placement.

The ROOM decay is doubled from the Studio's natural 0.36 s. On its own the treated studio is almost too dry to hear as "reverb", and 0.7 s is the small room musicians expect. The pre-delays grow with the size of the place, to keep the source clear of its reverb.

Lite has no dry/wet control, and doesn't need one. For a send/return, simply turn DISTANCE further out. Lite exposes seven parameters to the host — Place, a Size for each of the four places (only the current place's Size is shown), Distance and Bypass — so automation lanes stay simple. It has one program, "Default", and no preset menu.

Rule of thumb for Lite: vocals and lead instruments sit between CLOSE and the middle. Pads, choirs and backing parts sit from the middle toward FAR.

Signal flow

IN ─┬──────────────── DIRECT  (per-speaker mic pickup · distance level · air absorption) ───────────┐
    │                                                                                            ├─ OUT
    └─ INPUT BAND ─┬─ EARLY FIELD ─ MIC (polar patterns, per speaker) ─ tone ─ level ─────────────┤
                   │    ├ specular: image sources on the room polyhedron (order ≤ 3, Tunnel 4)   │
                   │    └ diffuse:  ray tracing "diffuse rain" (384 rays · 5 bounces)           │
                   │         └ part of the early field feeds the late network                    │
                   └─ PRE-DELAY ─ DIFFUSION ─ HIGH-DENSITY FDN ─ speaker route ─ tone ─ DYNAMICS ─ level ┘
                        16 lines · Hadamard · in-loop allpasses · 3-band damping · modulation
                        N inputs (own pre-delay, diffusers, Walsh input row per channel)
                        N decorrelated outputs (≤ 15 mains; LFE silent)

Specifications

Validation

Every release is checked objectively before listening tests:

• DSP test suite: 274 checks. FDN stability and resonance, RT60 per band, echo density, early-reflection geometry and timing, the distance model, tone, dynamics, click-free parameter changes, sample rates from 44.1 to 192 kHz, the microphone and immersive rendering, placement and rays.
• Apple auval (strict): passed for all three editions.
• pluginval at strictness 10: passed with 0 warnings for AU and VST3 of all three editions.
• Host harness (AU and VST3): layouts, 6 sample rates × 4 block sizes, irregular blocks, automation, bit-identical state restore, programs and the editor.

Known limitations (v1.1.4)

• The late field is statistical, not geometric. Position changes its level and onset, not its texture.
• Geometric acoustics does not model diffraction or low-frequency room modes. The low end of each space is shaped by its tuned late field.
• The immersive output is a channel bed, not Ambisonics or Atmos objects.
• The window opens at 1120 × 720 (Lite 560 × 460) and can be resized or zoomed from its edge, or from Window Zoom in the gear icon's utility panel; its aspect ratio stays fixed. User presets cannot be renamed or deleted from the interface.
• AAX is built and PACE Eden signed for all three editions but has had less testing in Pro Tools itself than the AU and VST3 builds.