World Labs showcased a preview of its interactive 3D scene generator

World Labs, the AI startup led by artificial intelligence scientist Fei-Fei Li, recently shared some preview outputs of its AI system, which can generate interactive 3D scenes from any 2D image. The startup calls this system its "first step towards spatial intelligence." World Labs' blog post lets users explore some of the 3D scenes generated by its AI, rendered right within their browser.

The visualizations are somewhat limited in their scope, triggering a warning that the user has reached a boundary within seconds of "moving forward". However, they are quite impressive in the range of effects the system can incorporate into its outputs, such as depth of field and dolly zoom effects. They are also unique in that, in some of them, users can interact with the objects, and passive effects like 'rustle' or 'waves' can be uniformly applied to the scene content. Although the featured scenes have a distinct video game, fantasy-like, or even cartoonish aspect, the simulations generally appear consistent and solid, with their contents strongly adhering to the laws of physics. These features do help avoid the not-quite-real effect sometimes present in AI-generated video.

This preview is a significant milestone for World Labs, which emerged from stealth a few months ago after silently securing $230 million from investors including Andreessen Horowitz, NEA, and Radical Ventures. In a recent panel at TechCrunch Disrupt 2024, Sound Ventures' Ashton Kutcher, Guy Oseary, and Effie Epstein spoke about their bets on the future of AI. There they also took the opportunity to confirm their investment in World Labs as part of their dedicated $265 million AI fund.