Draw with rules, then change them

Neural Drift starts with particles moving through a changing flow field. Their paths form an image over time. I made the rules adjustable so the experience is something to explore while it runs.

Particle count changes the density. Flow intensity and noise scale change the movement. Trail persistence controls how much of the previous motion remains visible. Each control keeps its numerical value beside it, connecting the setting to what appears on screen.

A random result is easy to find and easy to lose. The seed field gives the experiment a reference point. Previous and Next move through numbered seeds; Random Seed starts a different exploration.

The implementation seeds both the random positions and the noise field when it restarts the particles. Reset restores the original parameters. A seed identifies the starting conditions, while the animation continues to evolve frame by frame.

Particle speed determines the blend between two selected colors. A separate background control changes the field beneath them. Color becomes part of the behavior, rather than a filter applied after the image is made.

Underneath the interface, p5.js draws the particles. Layered Perlin noise guides the broader flow, while nearby particles influence one another. A spatial grid limits the neighbor search. This is procedural animation, not a trained neural network or a text-to-image service.