Towards physics-consistent and controllable generative worlds.
Generative models should model worlds—not just pixels.
Navigate with ← →, Space, or swipe · M for map · F for full screen
How can we build physics-consistent and controllable generative worlds?
Consistent across camera, time, and space. Controllable through conditions, goals, and constraints.
Four steps toward physics-consistent and controllable generative worlds.
Generative Photography
CVPR 2025 HighlightChange the camera, preserve the scene.
NewtonGen
ICLR 2026Generate motion that follows physical laws.
SeeU
CVPR 2026Generate the world beyond observed space-time.
OptiWorld
Under ReviewChoose a better future under constraints.
Control the camera, preserve the world.
Control focal length, bokeh, shutter speed, and color temperature while preserving scene identity.

Give video generation a sense of physics.
Learn Neural Newtonian Dynamics, then guide video generation with explicit physical states.

See beyond the observed time and space.
Lift sparse 2D observations into a continuous 4D world across unseen times and viewpoints.

Plan a better future, then render it.
Optimize goal-directed trajectories under physical constraints before rendering the future.
