PANews reported on December 3 that World Labs, a spatial intelligence company founded by Stanford University professor Fei-Fei Li, announced the development of an AI system that can generate a complete 3D world from a single input image and solve the problem of lack of control and consistency of existing generative AI models. The system can not only estimate 3D geometric shapes, but also fill in the unseen parts of the scene and generate new content, allowing users to freely explore 3D scenes in a game-like way.
The generated 3D scene supports real-time rendering in the browser, allowing simulation of 3D camera effects (such as shallow depth of field and dolly zoom). Compared with the generative model of pixel prediction, this system follows 3D geometric rules, the scene is consistent and supports dynamic interactions, such as modifying lighting, appearance and inserting new objects.
