Neural Cellular Automata: From Cells to Pixels
Summary
The Show HN article presents Neural Cellular Automata (NCAs) that generate high-resolution textures and shapes by coupling a coarse-grid NCA with a lightweight decoder (LPPN). The approach enables real-time, high-resolution outputs across 2D and 3D domains, including mesh textures, with interactive brush-based demos and GitHub code. The work, with authors from EPFL and Google Research, was showcased at SIGGRAPH 2026 and distributed via arXiv and open-source implementations.