Model Synthesis

Model synthesis, also known as wave function collapse, is a family of constraint-solving algorithms used in procedural generation, especially in video games. Given a small input example, such as a bitmap, the algorithm analyzes which local patterns occur in it and then generates new, larger output that reuses those same local patterns while satisfying the adjacency constraints they imply, progressively narrowing each output cell to a single definite value in a process compared to the collapse of a quantum superposition. The technique was first described by Paul Merrell in a 2007 conference paper under the name model synthesis, but it became widely known as wave function collapse after Maxim Gumin published an implementation under that name on GitHub in 2016, an implementation that also drew on ideas from convolutional neural networks and became the dominant variant used by game developers; games reported to use variants of the algorithm include Bad North, Townscaper and Caves of Qud. This description is adapted from Wikipedia contributors under CC BY-SA 4.0; changes were made. https://creativecommons.org/licenses/by-sa/4.0/