nine knobs.
one craft.
The Full edition opens the same DiffGFDN engine across nine knobs. Five small ones for time-domain shaping; four large ones for character and balance. Every knob renders identically — cream body, copper indicator — so you find them by label and position.
predelay — distance before the room answers
Row 1 · Range: 0 – 200 ms · Default: 12 ms
The gap between the dry attack and the first reflection. Short predelays glue the source to the room; long predelays separate the two so the dry transient stays clear in front of a long tail.
Studio-folk default sits at 8–18 ms — long enough for a vocal consonant to land before the room talks back, short enough not to feel unnatural. For drum bus, try 30–60 ms with Sentient decay; the snare hit hits twice.
early — level of first reflections
Row 1 · Range: 0 – 100 % · Default: 55 % · Pairs with: Body
Controls the level of the geometric early reflections — the first 4 to 12 distinct echoes that define the sense of place. Low values give a denser-feeling tail without spatial cues; high values make the room geometry obvious.
Pulled from the same DiffGFDN scattering matrix that the late field uses, so the early and late always agree about size.
body — mid-tail density
Row 1 · Range: 0 – 100 % · Default: 62 % · Pairs with: Decay
Density of the middle of the tail. Low Body keeps the tail thin and discrete (think early plate); high Body packs the modes together (think hall).
On vocals, Body is the difference between hearing individual reflections and hearing a wash. Most decisions about "is this reverb obvious" are decisions about Body.
decay — tail length
Row 1 · Range: 0.1 – 30 s · Default: 1.8 s
RT60 — the time the tail takes to drop 60 dB below the steady-state level. The exponential mapping is logarithmic to taste: small adjustments around 1 s feel detailed, big adjustments around 10 s feel cinematic.
Decay above 12 s engages a denormal-prevention noise floor — the engine keeps the network alive without bleeding inaudible noise back into your bus.
bloom — late reverb swell
Row 1 · Range: 0 – 100 % · Default: 30 % · Pairs with: Decay
The amount of swell the tail does after the dry has stopped. At zero, the tail is a clean exponential. At max, the tail rises before it falls — a slow inhale that adds drama to long-held notes.
Bloom is the parameter most responsive to the Brain. Loud transients grow more bloom; quiet sections stay dry. Automate it for breathing texture.
texture — algorithmic character of the diffusion
Row 2 · Range: 0 – 100 % · Default: 50 %
Selects between four topologies of the DiffGFDN scattering network — from clean (Hadamard-style, mathematical) to grainy (Householder with modulation, organic). The full range is a continuum, not four steps.
Texture above 70 introduces audible modulation in the late tail; useful on synthetic sources that benefit from a tape-style wobble. Below 30 the tail is forensically clean — good for orchestral material.
width — stereo width of the wet image
Row 2 · Range: 0 – 200 % · Default: 100 %
0 collapses the wet field to mono — useful when the source is centred and you want a halo without spread. 100 is the natural width of the network. 200 widens beyond the source for ambient effects.
Width is applied at the output, after the network — it doesn't change the dynamics of the early/late mix. Mono compatibility is preserved at any setting (the wet always sums down without phase cancellation).
tone — high-frequency damping in the tail
Row 2 · Range: 0 – 100 % · Default: 55 %
A frequency-dependent decay multiplier — controls how quickly the high end disappears. Low Tone gives a dark tail (cathedral, plate); high Tone gives a bright tail (chamber, drum room).
The damping is applied per delay line in the GFDN, so it scales naturally with Decay: longer tails get more damping per second of decay, which keeps the high end from being stuck.
mix — dry / wet balance
Row 2 · Range: 0 – 100 % · Default: 30 %
Linear crossfade with a -3 dB pan law. At 50% the wet is at unity. At 100% the dry is silent — useful when you want to send Arna on an effect bus and let the host handle the dry.
The dry path has a 0-sample latency; the wet path has the host-reported plugin latency, which is zero in Arna (no lookahead). So Mix automation is sample-accurate without comb filtering.
three buttons. one keyboard. one brain.
Spaces
A curated browser of presets organised by space type — Cathedral, Plate, Chamber, Hall, Spring, Drum Room, Vocal Booth, Cinematic. The selected preset shows a cream halo plus bold-cream text; everything else is a gradient of bgOrangeDark alpha.
Presets & A/B
The preset bar lets you snap to a saved setting. The A/B button on top of the slot reveals a second slot — Slot A is cream, Slot B is warm. Both colours live inside the warm palette so the comparison stays calm.
The Diff overlay uses the same two colours on the spectrum visualizer: cream for A, warm for B, with a copper meeting line where they agree.
Freeze
The only icy thing in the system. When engaged, the network feedback gain goes to unity — the tail is mathematically infinite. The button breathes at ~2 Hz so you remember it's on.
Each knob also gains a perpendicular white tick on its indicator while Freeze is active, so daltonic users have a shape signal — they don't depend on iceBlue alone.
MIDI keyboard
A 14-key strip across the bottom lets you play pitched chords through Arna without an external source. Useful for sound-design workflows, freeze-and-fly arrangements, and demonstrating the plugin without a session.
Active notes glow in vizGreen — the same colour as the visualizer waveform.