Everyday sounds — footsteps, scrapes, collisions — can be synthesised using the same Excitation × Vibration model. Physical parameters like surface material and contact force map directly to acoustic parameters.
Footstep synthesiser
A footstep has three simultaneous components: a low-frequency body impact (weight hitting the floor), a mid-frequency floor resonance (the floor material ringing), and a high-frequency surface texture (shoe sole on surface). Together they define the material and mass of the walker.
Footstep
Impact physics: sole mass hitting floor material. Every step varies slightly — add randomness to hear natural variation.
70 kg
Step speed
Footwear
Waveform
Scrape synthesiser
Scraping is a series of micro-collisions — the surface roughness causes rapid repeated excitation of the object's modes. Slower speed means fewer impacts per second; pressing harder excites higher-amplitude modes.
Surface scrape
Stochastic impact train through material resonance. The roughness of the surface and the hardness of the scraper shape the texture.
Medium
Medium
400 Hz
Waveform
Collision & impact
A collision is the simplest case: two objects, each with their own modes, each excited by the impact. The relative hardness of the objects shapes the contact time and therefore which modes get excited.
Object collision
Object 1
Object 2
Medium
Waveform
Why physical synthesis scales
Every step is unique. A recorded footstep sounds right once and repetitive the fifth time. A synthesised footstep varies naturally with each trigger — weight transfer, surface contact area, and shoe deformation all introduce randomness at the source, just like the real thing.
Parameters are physical. "Walker weight" directly scales the low-frequency impact component. "Surface hardness" changes the modal decay rate. These map to real quantities a game engine or physics simulation already tracks — the audio system can drive synthesis from physics state rather than from sound selection logic.
Scalability for interactive media. A game with 100 surface/object combinations would need 100 × N samples to cover all variations. Physical synthesis generates each combination at runtime from a small parameter set — one model, unlimited variation. This is the practical payoff of Cook's research for real-time interactive audio.