generative-art-algorithms

Generate an angle at each grid point using noise(x * scale, y * scale) * TWO_PI, then convert it to a vector with cos and sin. Particles sample the vector at their position each frame and apply it as a steering force, drawing trails as they move.

Fractal Brownian motion layers noise at increasing frequencies and decreasing amplitudes to create natural complexity. Use 4-8 octaves with lacunarity of 2.0 and gain of 0.5 for balanced terrain or cloud-like textures.

Call randomSeed(seed) and noiseSeed(seed) before generation so every run produces identical output. You can also hash a string into a numeric seed to tie reproducibility to a meaningful name.

Yes, the GLSL snippets use Shadertoy-compatible uniforms like iResolution and iTime, and work in any WebGL context including Three.js ShaderMaterial. They include SDF shapes, Voronoi, fBm noise, and cosine palettes.

Noise that looks too uniform means the scale is too large—decrease it for more variation. Chaotic noise means the scale is too small or octaves are too high; increase scale or reduce octaves for smoother results.

Perlin noise produces smooth continuous gradients ideal for terrain and flow fields, while Worley (cellular) noise measures distance to random feature points, creating cell-like patterns useful for organic textures like scales or cracked surfaces.