Spatial Intelligence and 3D World Generation
World Labs is a company founded by Feifei Li in 2024. Its focus is on the development of “spatial intelligence” – enabling AI to not only identify what is contained within images, but also to understand the structure of three-dimensional spaces in the same way that humans do, and to be able to reason and act within those spaces.
At present, our main product for external use is Marble. By submitting an image or a description, users can create a 3D scene that allows for free movement and adjustment of the viewing angle – rather than a fixed-position video. This represents a completely different approach compared to mainstream video generation models. While those models produce sequences of pixels, World Labs aims to create interactive environments in themselves.
The target audience is quite clear: teams involved in creating white-box levels for games, pre-production visualizations for films and television, architectural and spatial design projects, as well as robot simulation training. Researchers interested in generative 3D technologies would also find it useful. However, ordinary content creators may find the learning curve relatively steep; currently, this tool is not yet suitable for creating short videos.
The official website provides an online trial interface, as well as APIs and research resources for developers. The product is still in the midst of rapid development, and at present, the scale of the generated scenarios and the level of detail accuracy represent its main constraints.