Summary
WorldCrafter is a video world model designed for interactive, streaming exploration of dynamic environments with strong long-horizon and multi-view consistency. Its core idea is a camera-queryable implicit 3D-aware memory: the requested viewpoint determines how multi-view evidence from prior observations is compressed into the video generator's limited token budget. A memory encoder and a pose-conditioned readout module, trained jointly with the video generator, integrate historical observations into a fixed set of target-view tokens before denoising, without relying on explicit depth-based correspondences. Combined with recent temporal context and few-step distillation, the system supports streaming scene exploration starting from a single input image or a text prompt. Experiments on both static and dynamic scenes show significant improvements in long-term consistency and camera control accuracy while preserving visual quality during minute-scale exploration. The paper (arXiv: 2609.24984) is authored by Wangbo Yu, Kunhao Liu, Wenbo Hu, and colleagues from a team including Ying Shan.
Paper Overview
Field: Computer Vision (CV)
Authors: Wangbo Yu, Kunhao Liu, Wenbo Hu, Shenghai Yuan, Chaoran Feng, Haiyang Zhou, Yukun Huang, Yiran Wang, Wang Zhao, Yingmin Luo, Ying Shan
Published: 2026-09-21
arXiv: 2609.24984
Abstract
Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. The authors present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory to address this.
The key insight: 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 a 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 a text prompt.
Results
- Experiments on both static and dynamic scenes
- Significant improvements in long-horizon consistency and camera control accuracy
- Maintains visual quality during minute-scale exploration
---
*Auto-collected on 2026-09-23*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178635098