Loading paper
MAGREF: Masked Guidance for Any-Reference Video Generation with Subject Disentanglement | Tomesphere