Open-Sora

“A frozen world model pψ, adapted from Open-Sora [Zheng et al., 2024], predicts action-conditioned multi-view futures from wrist and third-person observations.”