Signal intelligence // generative DSP ecosystem

Signal Garden.

A browser-native instrument that grows rather than loops. Granular memory, modal resonance, harmonic fields, sparse events, spectral processing, and feedback weather evolve on the audio thread while the interface exposes only the controls needed to shape the ecology.

Living signal engine

Start in Autonomous mode and let the ecosystem develop. Switch to Perform when you want direct control. Use headphones at a conservative level.

idle
50%

Final listening level after makeup gain and safety limiting.

Mode

The garden moves through long-form states by itself. Your gestures still influence it without taking over.

Bright resonant droplets, pentatonic gravity, high spectral dust, and slowly widening space.

70% ambient
ambientexperimental

Biases the autonomous engine toward spacious continuity or stranger, less predictable behaviour.

drag to attract grains vertical = pitch field horizontal = stereo field
DORMANT field // glass pentatonic motion // drifting
Current ecology DORMANT

The garden is holding energy in long grains and low-density resonance before the next growth cycle.

harmonic bed modal bloom rare events air FM colour subharmonic
RMS -∞ dBFS peak -∞ dBFS grains 0 voices 48 DSP 0% sample rate not started
Shape the ecosystem Cross-pollination, macro movement, weather, spectral veil
Harmonic genetics

Cross-pollinate

Glass Rain
0.00
0.14

Blend changes the harmonic vocabulary and motion genetics. Mutation introduces bounded deterministic variation while the seed keeps the result reproducible.

Performance field

Macro movement

volatile calm sparse dense

density 0.50 // volatility 0.50

One gesture moves several parameters together. The individual DSP controls remain below for inspection and fine adjustment.

Space and transformation

Environment

Feedback weather

Balanced feedback with restrained modulation and a clear stereo field.

Spectral veil

The granular signal passes directly into the feedback network.

Source and capture Use your own audio, record the output, share deterministic states
Local source

Feed the garden

Capture

Record this ecology

Recording uses the browser media pipeline after the safety limiter and master-volume stage.

Determinism

Seed and share

The URL can recreate the ecosystem, seed, character, mode, weather, and spectral state.

Advanced DSP Fine controls and live engine telemetry
0.78
0.07 s
0.82
0.88
0.12
0.28
0.92
0.42
0.72
0.96
0.34
0.18

Procedural memory

Seven synthesis layers continuously write new material into circular memory: harmonic partials, filtered noise, FM colour, subharmonics, air, sparse events, and a tuned modal resonator bank. Granular voices read that evolving history instead of repeating a fixed oscillator drone.

Four clocks of change

Micro, phrase, ecosystem, and epoch modulation operate at different timescales. Short motion changes individual grains; longer processes reshape density, harmonic fields, event probability, and synthesis-layer balance over minutes.

Lifecycle engine

Autonomous mode moves through Dormant, Germination, Bloom, Migration, Storm, Decay, and Regeneration states. Each state changes the probability and intensity of sonic behaviours without replacing the underlying ecosystem.

Browser to game engine

The same ownership boundary in UE5 MetaSounds

BrowserMetaSounds
AudioWorkletProcessorMetaSound operator graph executing in the audio renderer
Seeded lifecycle and event schedulersDeterministic trigger-rate generators and state machines
Circular synthesis memory and interpolated grain readsWave assets or a custom granulator operator with explicit voice state
Modal resonator bank and harmonic fieldsResonator operators, filters, and data-driven tuning tables
Smoothed AudioParam intentGraph inputs with interpolation before audio-rate destinations
Bounded feedback and safety limitingExplicit feedback clamps, dynamics control, and output protection

Generative systems, not random playback

The reusable idea is layered ownership. The interface chooses intent and exposes state; the audio renderer owns timing, voices, envelopes, resonators, feedback memory, deterministic evolution, and safety. A seed can reproduce the rules of an ecology without forcing every rendered sample to repeat exactly.

The same distinction matters in games. Gameplay code should request a mood or state transition; the audio system should decide how that state becomes sound inside a bounded real-time budget.