On August 17, HiDream.ai officially launched HiDream-O1-World, a native all-modal interactive world model that delivers three core capabilities: roaming, editing, and interaction. The system accepts multimodal inputs including text, images, and interactive commands, allowing users to generate—with a single click—complete worlds that are spatiotemporally consistent, physically plausible, and capable of long-horizon simulation.
Under the hood, HiDream-O1-World is powered by HiDream’s self-developed native all-modal (UiT) architecture. According to the company, UiT has achieved critical breakthroughs in both spatiotemporal consistency and physical consistency, enabling the model to maintain coherent scene dynamics and realistic behavior over extended sequences.
