World Labs, an AI start-up founded by Fei-Fei Li, a renowned computer scientist and AI pioneer, has unveiled a new system to generate interactable 3D 'Worlds' from 2D image inputs. The system is part of an emerging trend of AI world models that can estimate an environment and build a representation of it.
The company has some examples on its website that show off the technology. These can be explored using simple mouse and keyboard controls, but explore sounds a little generous here. In reality, the interaction is limited, and it only takes a few steps in any direction to see an "out-of-bounds" message.
Are you a techie who knows how to write? Then join our Team! Wanted:
- News Writer (Romania based)
Details here
The world feels stuck in the uncanny valley with a flat aesthetic that isn't entirely 3D. Nevertheless, it is impressive for an "early preview." Once a scene is generated, the system can add fully controllable effects like depth-of-field or more cinematic camera techniques like dolly zooms.
World Labs says the generated scene offers real-time control and obeys the basic rules of 3D geometry. World Labs says it is "hard at work improving the size and fidelity of our generated worlds, and experimenting with new ways for users to interact with them."