Machines that learn to render.

The next generation of visual media will be generated, not captured.

We build image models that understand light, material, and composition + every layer required to train them.

Read the research.

Notes from the lab on what's in progress — open as we go.