Sonifold: audio-driven nodal surfaces on 3D meshes

In 1787 Ernst Chladni scattered sand on vibrating plates and revealed the hidden geometry of sound. Two centuries on, a direct counterpart for digital audio on closed three-dimensional surfaces remains less developed. Sonifold extends Chladni's construction to such surfaces by mapping audio spectra to Laplace–Beltrami eigenfunctions and extracting the nodal sets of the resulting scalar fields. Because the eigenbasis is intrinsic to the geometry, the same signal yields different patterns on different shapes, making shape a geometry-dependent spectral filter. Across nine meshes and seven stimuli, a seven-point genus sequence shows a strong negative correlation between genus and β0 (Spearman ρ≈−0.93), opposite to the naive Weyl-density prediction. Visual complexity tracks the eigenstructure, not topology alone. We frame this principle as ‘shape selection is filter selection’, a generative parameter that practitioners can sequence over time, and accompany the paper with an interactive web application.