Click the canvas to drop sites (seeds). Every pixel is colored by the site closest to it, so the boundaries you see are the perpendicular bisectors between neighboring sites — that's a Voronoi diagram. Try different distance metrics: Manhattan bends cells into rectangular shapes, Chebyshev makes square cells, and Euclidean gives the classic convex polygonal cells used in mesh generation, nearest-neighbor lookups, and 2D coverage maps.
Tip: click empty space to add a site, click on a site to drag it, right-click (or long-press) a site to remove it.
Given a set of points in a plane (called sites), the Voronoi diagram partitions the plane into regions where every point inside a region is closer to its associated site than to any other site. The construction is named after Russian mathematician Georgy Voronoy (1908), though the idea goes back to Descartes (1644) and Dirichlet (1850, hence "Dirichlet tessellation"). It's the dual of the Delaunay triangulation and shows up everywhere: cellular network coverage, robot path planning, generative art, k-nearest-neighbor classification, fracture patterns in geology, and the textures of giraffe spots and dried mud. The optimal Fortune's sweep-line algorithm runs in O(n log n), but for visualization a brute-force "each pixel finds its nearest site" works fine up to a few hundred sites.