With Atlas, we introduce the idea of “spatial context”. Similar to an LLM, Atlas first encodes its inputs into a context, then generates outputs conditioned on the context. However, in Atlas, each image is grounded at a 3D position in space; this forms a ‘spatial context’. Managing this spatial context unlocks new kinds of creative control. For example, you can place two totally unrelated images in the context, position them anywhere in 3D space, and Atlas generates a world that smoothly…