When I heard Debussy's Clair de Lune, a Monet surfaced in my mind. Similar things always happen.
I've always wondered what it is — what features make me feel that a certain piece of music looks like a certain painting. It's a personal question; I'm chasing the mechanism behind my perception.
Correspondence is my way of reaching for that question. It maps the colors in an image into sound — not to make a product, but to build an instrument that can translate, and to see whether a painting, once it becomes sound, comes out feeling right to me.
More seriously: Correspondence is my attempt to reverse-engineer cross-modal perception by building a parameterized instrument for testing where, and why, sound and image start to feel like each other.
how it works
Correspondence is designed to read pixel values from an image under the user's cursor and synthesize music from them in real time — hue mapped to pitch, saturation to dynamics, brightness to tempo. The synthesis core is being built in Rust and C++, meant to compile to WebAssembly and output through the Web Audio API, with every note coming from oscillators — no samples.
The engine is currently being built. A playable demo and the full technical writeup will be published here when it runs.