Sound & Interaction — Sonification · AR · Physical Modeling · Neural Audio

↗ Repository

AquiFuturo

Augmented reality experience that sonifies tree morphology — translating a Coast Live Oak's canopy and root structure into spatial sound through physical modeling and neural audio synthesis.

UnityARKitPythonRAVEC#Blender

Preview

How it works

3D Scan
(Photogrammetry)
Roots Simulation
(Rhizomorph)
Skeleton graph
Canopy
  • Modal Synthesis
  • Mass-Spring
  • 96 Modes
Roots
  • RAVE Decoder
  • Latent Space
  • 16 Dims
Unity AR App
(iOS / ARKit)

Specs

PlatformiOS (ARKit)
EngineUnity + AR Foundation
Audio tracks9 (5 canopy + 4 roots)
Canopy synthModal (96 modes)
Root synthRAVE (16-dim latent)
PipelinePython + Blender
Sample rate48 kHz / 24-bit

Concept

AquiFuturo is an augmented reality experience that sonifies a Coast Live Oak. The canopy skeleton is sonified via modal synthesis mirroring structural hierarchy; the root system via RAVE latent-space navigation trained on water sounds. The visitor approaches the tree, places the root system beneath it in AR, and listens to the morphology.

Technical Detail

  • —Canopy sonification via modal synthesis: 267-node mass-spring network with per-class stiffness multipliers stratifying the spectrum across five structural classes (trunk base 20–120 Hz to terminal tips 1,500–8,000 Hz). Eigenvalue decomposition extracts 96 natural frequencies and mode shapes.
  • —Root sonification via RAVE latent-space navigation: 8 morphological features per node mapped into a 16-dim latent space via PCA. A periodic cubic spline connects TSP tour nodes into a smooth trajectory, decoded in overlapping chunks with crossfades.
  • —AR app on iOS via Unity + AR Foundation + ARKit. Nine audio tracks: five canopy layers with orientation-modulated filtering and gain, four RAVE root-zone clips triggered by tap interaction.

Learnings

  • —Modal synthesis encodes morphological hierarchy into sound — trunk produces sub-bass, terminal tips produce brightness. But the mapping is not self-evident without interactive tools.
  • —RAVE opens sonic spaces beyond deterministic physical models, though the output is shaped by training data as much as by geometry.
  • —The pipeline generalizes — any tree with a photogrammetric scan could generate its own soundscape.

PROJECTS

Sound & Interaction · Data & Science