Adobe Research video world model that builds navigable interactive worlds from an image or video at 16 FPS.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Wonder is a research video world model from Adobe Research and Johns Hopkins University. Given an image or a video, it generates a persistent interactive world that a user can navigate by moving the camera, revealing unseen regions and returning to previously seen areas, with constant latency for up to one minute.
The system combines three design choices. Camera translation and rotation are turned into dense, frame-aligned visual evidence by rendering a synthetic camera space with a 3D scaffold and environment map. Full-fidelity history KV caches are kept and relevant entries are selected through sparse attention to maintain long-horizon memory. Training uses a Mixture-of-Students design and a GAN Control Regularization step to reduce camera drift during distillation. Rollouts run at 16 FPS, and the project page demos cover game, cartoon and real-world scenes, as well as video-to-video re-rendering with new camera paths.
At the time of this entry, the project page is a research demo. Code and models are listed as coming soon, and no public product, API or pricing is described. A reviewer noted visible noise and artifacts in the output.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Video world model that creates interactive worlds you navigate in real time with keys. It also accepts an input video to re-render with free camera movement. Quality shows noise and artifacts, and code and models are listed as coming soon.
Closest neighbours by meaning, not just by category.
Model · Runway
Runway's real-time interactive world model generating 720p/24fps video with audio, steered by text actions and camera input.
Model
World model that turns a few context images into an editable, physics-simulated 3D world rendered by a video model.
Model
Real-time interactive world model with compressed memory and hybrid action and semantic control, built on Wan 2.2.
Model
Research framework that turns the Minimax H3 video generator into an interactive world model controlled by key presses.
Model
14B autoregressive world model that generates interactive, camera-controlled video worlds in 3 denoising steps.
Model
Interactive autoregressive world model generating minute-long playable video with real-time camera control and prompt events.
