WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory
WorldCrafter is a video world model with implicit 3D-aware memory enabling long-horizon consistent scene exploration from a single image or prompt.
WorldCrafter learns a camera-queryable implicit 3D-aware memory so video generation respects prior observations over long horizons and across viewpoints. A memory encoder and pose-conditioned readout compress multi-view evidence into view-specific tokens before denoising, without explicit depth-based correspondences. Combined with temporal context and few-step distillation, it enables streaming scene exploration from a single input image or text prompt, with substantial gains in long-horizon consistency and camera-control accuracy during minute-scale exploration.
- Camera-queryable implicit 3D-aware memory conditions the video generator on history.
- Viewpoint shapes how multi-view evidence is compressed into a fixed token budget.
- Enables minute-scale streaming exploration from one image or text prompt.
Full article136 words · extracted from arxiv.org · click to collapse
Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. We present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory for this purpose. The key insight is to let the requested viewpoint shape how multi-view evidence is compressed into the video generator's limited token budget. Trained jointly with the video generator, a memory encoder and pose-conditioned readout module integrate historical observations into a fixed set of target view-specific tokens before denoising, without explicit depth-based correspondences. By combining this memory with recent temporal context and few-step distillation, WorldCrafter enables streaming scene exploration from a single input image or text prompt. Experiments across static and dynamic scenes show substantial gains in long-horizon consistency and camera-control accuracy while preserving visual quality during minute-scale exploration.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.24984