Adobe and Johns Hopkins unveil Wonder, a real-time controllable video world model
Adobe Research and Johns Hopkins University unveiled Wonder, a real-time, camera-controllable general video world model. It turns a single image or a video into an explorable "playable world" where users can move the camera in real time, reveal unseen areas beyond the frame, and revisit previous viewpoints—exploring generated content as if operating a real camera.
Confirmed
- The model supports 6-DoF camera control, enabling continuous interaction within the generated world.
- Video generation runs at 16 FPS, with visual consistency maintained for over a minute.
- Interaction latency stays constant at 0.5 seconds, delivering a real-time exploration experience.
- The system relies on system-level co-design across control representation, memory mechanism, and training strategy.
Why it matters
- Wonder moves beyond static images and one-way video generation, turning input into a dynamic, freely controllable playable world that opens a new technical path for games, film, and 3D content creation.
2026-07-29 ~ 2026-07-31 · 5 related posts
Primary sources
- [source] Adobe’s Wonder turns images into a camera-controllable video world model — adobe · 2026-07-29
- Wonder: Real-Time Camera-Controllable World Model at 16 FPS — qixing_huang · 2026-07-30
- Wonder: Real-Time Interactive 3D World Model from Adobe and Johns Hopkins University — yuyinzhou_cs · 2026-07-30
- Adobe's Wonder: Real-Time Interactive Video World Model with Constant Latency — aigclink · 2026-07-31
1 near-duplicate retellings: aigclink