InsertAnywhere: Geometrically Grounded and Optics-Aware Video Object Insertion

Hoiyeong Jin, Hyojin Jang, Junha Hyung +6

Recent advances in diffusion models have enabled impressive video editing capabilities, yet production-grade Video Object Insertion (VOI) remains challenging due to inadequate 4D s…