HiDream-O1-World: New AI Model Creates Explorable 3D Worlds

HiDream.ai has launched HiDream-O1-World, a new interactive world model. This model allows users to create full, explorable 3D worlds from just a text prompt, an image, or interactive controls. Here's the thing: it combines roaming, editing, and interaction capabilities. It's built on HiDream.ai’s self-developed UiT architecture, which addresses challenges like spatiotemporal and physical consistency. This means that as cameras move, scenes remain stable, and objects don't vanish or deform. Collisions and gravitational responses follow real-world logic. What's interesting is that the model topped the WBench leaderboard, scoring 80.9 on the Navi sub-leaderboard and 88.0 on Consistency. It also ranked first in the Physical dimension at 73.3. This performance sets a new standard for interactive world models. The model lets users generate diverse interactive worlds from a short description or a single image. For example, uploading a photo of a room can rapidly construct a high-precision digital twin. Users can explore these worlds in first-person or third-person mode, and even trigger environmental events like rainfall. This matters because it pushes the boundaries of AI's understanding and reconstruction of the physical world.