Our method reconstructs 3D worlds from video diffusion models using non-rigid alignment to resolve inherent 3D inconsistencies in the generated sequences.
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).