Video generation models as world simulation engines
This article explains how Sora, a text-conditional video generation model, creates minute-long high-fidelity videos. It covers how training on videos and images of variable formats expands visual know…
