World Labs reveals persistent 3D scenes you can actually stay in

If you've played with generative 3D or early world models, you know the frustration: things look fine for a few seconds, and then the scene either drifts into a hallucinated mess or boots you out.

World Labs just announced that its world model now generates persistent 3D scenes. According to the team, users can build on top of these environments and stay inside as long as they want, even if that stretches way past the typical 60-second mark.

No more 60-second limit

Most current video and spatial models struggle to keep physics and spatial consistency intact across time. World Labs is tackling that bottleneck directly by outputting environments that hold their shape indefinitely instead of dissolving.

The ability to "build on top" of these scenes is the key detail here. It suggests that these environments aren't just baked meshes or transient video clips, but functional spaces that accept ongoing user inputs.

Why it matters

Video generators are impressive, but you can't walk inside a flat video file or change the furniture. By introducing persistence, World Labs is steering generative AI away from passive media toward interactive game engines and real-time simulation.

We think scene permanence is the single biggest hurdle world models have to clear before they can be taken seriously by developers. A scene you can stay in indefinitely is the difference between a flashy demo and actual software infrastructure.

That said, details remain scarce. We still need to see how much computational heavy-lifting this requires and how far user interaction can actually go before things start to break.