Sound becomes geometry.
Explore how frequency, material and phase become visible structure. An independent study inspired by AcusMagic FIGURA.
Modal plate physics. Particle motion is a visual transport approximation. Tap the plate to move the excitation point.
Research record — mechanism, reconstruction and Spacetime fit
Reconstruct the mechanism.
The supplied post announces FIGURA by AcusMagic. Its public description exposes enough of the architecture to build an independent study. No FIGURA source code was inspected or copied; its internal implementation remains unverified.
| Stage | Published behavior | This reconstruction |
|---|---|---|
| Listen | Up to 32 partials; 170–190 ms history; 20 Hz–12 kHz; nearby tones merge at about 20 Hz. | Synthetic tones or local audio. An 8,192-sample browser FFT, strongest 8 local peaks, ≥30 Hz spacing. Window duration depends on sample rate: 186 ms at 44.1 kHz, 171 ms at 48 kHz. Updates about 5 times per second. |
| Sand | Modal surface response. Studio membrane with sheet stiffness; Lab free-edge plate in physical units. Default rig draws up to 24 modes. | 49 analytical modes of a simply supported square Kirchhoff–Love plate. Aluminum, 1 mm thick. Local-audio partials above the highest basis frequency are excluded; near-cutoff response still requires convergence testing. Distinct boundary conditions from FIGURA and the circular plate shown in the post. |
| Liquid | Faraday-wave model; subharmonic response; capillary–gravity dispersion. Rendering adds reflections and caustics. | Dispersion-derived wavelength and prescribed square, stripe or hexagonal fields. Weak-viscosity damping changes display amplitude. Local audio drives this view only from peaks between 20 and 160 Hz. No instability solver, fluid transport, vessel boundary or optical power model. |
| Phase | Stereo waveform drawn as a curve; L on X, R on Y; optional 3D orbit. | 2D analytical Lissajous curve or direct stereo samples from a decoded local audio file. Mono files form a diagonal. |
| Render | GPU particle and field rendering; native graphics; video export. | Browser Canvas, 7,000 particles, 112² field grid. No dependency, network request or native plugin needed. |
Sand: modal response
Complex amplitudes preserve interference between modes at one frequency. Energies are averaged across separate input frequencies; close-tone beating is not modeled. Gradient-driven grains are a visual heuristic, not contact mechanics. Response/static is the RMS displacement relative to the same load at zero frequency; field color is spatially normalized.
Liquid and phase
Liquid constants: ρ = 1,000 kg/m³; σ = 0.072 N/m; g = 9.80665 m/s²; circular field of view 80 mm. The γ approximation neglects boundary damping and becomes less reliable at high viscosity. Relief shading is illustrative. Display time is slowed explicitly to avoid implying a camera measurement.
Assumptions that stay visible
The plate uses E = 69 GPa, ρ = 2,700 kg/m³, ν = 0.33 and h = 1 mm. The initial drive is off-center at (0.31a, 0.43a); a centered source would suppress even-index modes. Particles are confined to the interior margin to prevent the supported edge from absorbing every grain. Scatter restores the same seeded initial distribution. Fixed simulation steps make the synthetic study repeatable on the same runtime; file analysis and floating-point rendering can vary across browsers. The canvas is normalized for legibility: brightness does not represent measured displacement, energy, confidence or optical power.
What Spacetime should take
| Reusable element | Research use | Evidence gate |
|---|---|---|
| Spectral analysis | Explore measured vibration or acoustic signals; connect frequency components to a declared physical model. | Calibrated data, units, sample clock, transfer function and independent reference. |
| Fields as the interface | Show a spatial field directly, with a probe for source, timestamp, model and uncertainty. | Keep uncertainty separate from visual grain, glow and aesthetic noise. |
| Recipe and replay | Capture model version, seed, settings, drive and provenance so the same synthetic state can be reviewed. | Exact replay additionally requires the input samples, timestamps and environment. This export does not embed local audio. |
| Assure integration | Candidate visual layer for research and sensor diagnostics. | No positioning, integrity, spoofing resistance or navigation claim follows from this demonstration. |
Critical path to a measured research result
- Reference measurement. Use a declared plate geometry and mounting, a calibrated drive and a camera/accelerometer sweep. Match the whole system, not just one frequency.
- Model validation. Check resonant frequencies and shapes, damping bandwidth, boundary conditions and convergence at higher mode count. Compare against independent experimental data before claiming agreement.
- Input validation. Use recorded tones/chirps to measure peak error, multi-tone separation and end-to-end latency. The current FFT is an instrument sketch.
- Runtime integration. Move field and particles to WebGPU only if profiling warrants it. Add source identifiers, clock domains, unit conventions, model revision and probe values to Spacetime's shared research state.
Verification record · October 21, 2026
Eleven numerical and interaction checks passed in a JavaScript runtime with Skia canvas rendering: analytical resonance, inverse-square size scaling, dispersion residual, particle settling, seeded repeatability, finite fields at parameter extremes, field toggle, liquid controls, phase controls, recipe round trip and invalid-recipe rejection. Mean normalized particle potential fell from 0.2462 to 0.00980 after 480 fixed steps. These checks verify implementation consistency; they do not validate physical accuracy. The three canvas views were rendered and inspected. Full browser layout, real audio decoding/playback, mobile performance and experimental agreement remain unverified.
Primary references
- AcusMagic — FIGURA: architecture, listening limits, material models, rendering and license details. Accessed October 21, 2026.
- AcusMagic — Sonicrama: shared engine, plate-shape approximations, liquid spatial-frequency compression and declared limits.
- Official Sonicrama 2.4.0 manual: reduced physical fields and constructed pattern families are separate approaches; neither is a full 3D fluid solver.
- University of Colorado — Chladni Figures: physical sand collection at nodes; effects of geometry and frequency.
- Tuan et al., JASA 137, 2113–2123 (2015): experimental plate-resonance study cited by AcusMagic. No experimental agreement is claimed for this reconstruction.
Status: independent browser reconstruction of published principles. Original FIGURA product claims are attributed to its creator; the study is not affiliated with AcusMagic. Physics validation requires measurement. Research prototype, not an Assure navigation component.