Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Move Your Bubbles!
Learn how Neural Cellular Automata turn static images into naturally moving textures in 2D and 3D, with interactive demos and low‑parameter models.
Today, fancy images are everywhere. Have you wondered what it looks like if they move naturally? In this project, I will introduce a bio-inspired tool, Neural Cellular Automata (NCA), for dynamic texture synthesis in the 2D and 3D domains. By modeling each pixel as an intelligent agent that can communicate with each other, NCA generates naturally moving videos given static target images. Apart from the emergent dynamics, NCA enjoys various test-time properties such as robustness to perturbation, motion control, and grafting. It can be used for game development, special effect creation, and creative arts. Different from current deep learning models that have hundreds of millions of parameters, NCA typically has less than 10000 parameters, rendering it easy to deploy on edge devices such as your phone!
DyNCA utilizes Neural Cellular Automata for real-time, controllable dynamic texture synthesis.
MeshNCA synthesizes dynamic PBR textures on 3D meshes via WebGL.
Compose Email
Loading recent emails...