AI · Shaders
"Slow purple plasma reacting to the bass." "Tunnel of neon waveforms." The generator writes a complete fragment shader from your prompt and drops it in the library as a regular shader source — audio-routed dials already wired, ready for a clip slot.
Every generated shader exposes one to four user dials, and each ships with a default audio binding chosen to fit the visual — frequency bands from sub to brilliance, beat onsets, sweeps, or random-on-beat jumps. The shader moves with the music the moment you load it; the dials stay yours to re-bind, ride on MIDI, or automate.
The server checks every generation before it reaches your library — the code has to be well-formed, the dials have to actually drive the visual, the audio uniforms have to be used correctly. A generation that fails validation never lands broken; it's rejected and the tokens come straight back.
The output is plain GLSL in the same editor you'd hand-write in — open it, tweak it, hot-reload it live. Generate ten variations for the price of a coffee, keep the two you love, tune them by hand, and publish the results to the marketplace under your name.
