A self-contained generative stereo instrument. It never repeats: every run invents its own key, chord progression and timbre. Nothing to install, works offline. Tip: hover any control for a short explanation.
The two markers are your left and right speakers. Drag to conduct the sound in real time: sideways = pan, up = brighter + more echo, distance from the centre = decay / length. Tap anywhere to play a chord at that position and pitch. The glowing dot shows the real stereo position of the output (it only moves when the sound actually moves).
start plays and stops. new take keeps your settings but generates a fresh run — new key, progression and timbre. random dices every setting for a surprise, while keeping the chosen character and its environment and always keeping an audible foundation.
Twelve characters — and each one is a portrait of a single sound role. Hauke is one deep spectral drone; Wiebke lets the background chords bloom; Klaas throws modal chord stabs into harbour fog; Telse is the acid line that bites; Antje sings a lyrical wavetable lead over water; Onno hammers iron modal percussion; Momme puts the speaking voice on the pulpit; Gesa vaults a choir under cathedral stone; Frauke whispers close to the ear beneath an aurora; Broder lets the place itself play (a rainforest of calls and beds); Swantje is pure sky shimmer; Eiko lives between stations in shortwave static. Pick a character to hear what one role can carry — then push its layer further, or lock it and travel on.
The stack of rectangles on the right is the heart of the surface: one rectangle per layer, each drawing the live waveform of exactly that layer's bus in its role colour — a muted layer shows a dim flat line. The name of the assigned synth sits in the corner. Click a rectangle and its settings unfold beneath it (synth choice and all parameters); clicking another closes the first. The small ○ button fixes the layer: a locked layer (●, dashed amber frame) survives Random, New take, character switches and Evolution — including the patterns that feed it (the lead keeps its motif and rhythm, percussion its pattern, the voice its character and phrases). Only the global key keeps moving so a locked layer stays in tune.
Warmth low-end body · Brightness high-frequency openness · Resonance filter emphasis · Saturation analog drive.
Movement amount/speed of stereo motion · Width stereo width · Reverb convolution reverb with four generated impulse responses — hall (soft), plate (short, bright), cavern (long, dark), air (open, early reflections). Each character brings its own room; changes crossfade without clicks, and the dice may swap the room · Echo tempo-synced ping-pong delay.
Harmony chord richness (extensions, suspensions, tension) · Density how busy the texture is · Events a live master for how MANY sound events happen — fade from desert (0.25×) to dense city (3×) · Dominance tilts between background (drone, pads, ambient beds) and foreground (lead, shots, percussion, calls) · Length stretches or shortens the tonal events themselves · Evolution how fast patterns mutate · Chord change chord length in bars · Breath subtle tempo drift · Chaos a Lorenz attractor slowly bends echo time (tape drift), the acid position and voice levels — continuous movement, never stepped.
Tempo in BPM (also sets the echo timing). Register is the base octave. Scale picks the musical mode and overrides the character’s own scale.
Voice and ambience each offer three forms, chosen like any synth. The voice speaks (formant speech, rhythmic syllables), sings (choir — sustained vowel pads on the chord tones) or breathes (whisper — the same syllable patterns as voiceless shaped noise). The ambience is a concrete place (field recording: calls and beds of the character), an abstract sky (aurora: slowly drifting partials and rare glints) or the ether itself (shortwave: wandering noise bands, heterodyne whistles, static). Characters carry their own forms — Frauke whispers under an aurora, Onno hums in shortwave static, Gesa sings.
Under the circle sit three pads — live gestures, not bar-quantized. TIME freezes and stutters the whole mix: X sets the slice (1 bar → 1/32), Y the freeze amount, a tap catches the current bar and lets it ring. WORLD rides the event macros: X density (desert → city), Y length, tap flips foreground/background dominance for one bar. SPACE uses both axes twice: X morphs the room toward its opposite and sets the speed of a global flanger, Y drives the flanger depth and the echo feedback; tap fires a one-moment wet burst. WORLD and SPACE glide back to where they started when you let go — TIME releases immediately. Locking whole panels: click the Timbre, Movement or Grab delay heading to lock every control in it against Random — single controls still lock by clicking their name.
On a monitor the instrument arranges itself in three columns — touch field left, layers in the middle, character and setup right — sized to fit one screen without scrolling; rarely used panels (Scale, Performance, Export, Colors, Grab delay) collapse behind their ▸ headers. Every change you make commits on the next bar line: while queued the control shows a dashed outline, at the moment it takes effect it briefly glows. Switching characters while playing is the same idea writ large — the staged crossing walks one layer per bar into the new world.
Eleven synthesis engines: Analog (detuned saw/square/tri with sub), Additive, FM, Formant, Ring, Fold (wavefolder), Granular (grain clouds), Spectral (drifting partial wash), Modal (struck bars, bells, plates, wood), Wavetable (morphs between frames), Noise (filtered noise — as percussion the classic drum synthesis, elsewhere a breathing texture). Each role plays exactly one synth, so the structure itself keeps the CPU load and the density in check.
A tempo-synced ping-pong echo that grabs single chord notes and lets them bounce through the stereo field — Mix, Repeats, Time (beat divisions), Rate (how often it grabs) and Width.
Every change you make lands musically: it is queued and applied at the next bar line — the marker shows the countdown. Set mode plays a whole set for you: every N bars (2–64) it moves to a new character or a fresh take, crossfaded.
Export records the master output — exactly what you hear — and saves one stereo WAV mixdown. Stop cancels a running recording.
The acid line may spiral: an accumulating transposition shifts each pattern pass by a small interval and folds back after an octave. Accents follow a Gaussian spread instead of hard randomness, and percussion and ambience positions wander through the stereo field as Brownian walks — the next value stays close to the last instead of jumping.
Always on and cannot be switched off: a DC blocker, a coupled look-ahead limiter (−1 dBFS, stereo image preserved), a hard clamp, a NaN/Inf guard and denormal flushing. Whatever you dial in, the output stays within safe bounds without clicks.
Choose the interface palette at the bottom — Deep Space, Amber, Phosphor, Slate or Magenta.
generate possibilities · not presets