WorldDirector Decouples Semantic Motion from Visual Rendering in Video Models

WorldDirector uses an LLM to coordinate 3D trajectories and camera movements, separating motion orchestration from pixel generation. This decoupling ensures physical logic and object identity persistence even when entities re-enter a scene after being out of view.