Preprint
Aug 2026
ScaleVid: Geometry-Aware Video Object Scaling with Mesh-Free Inference
This work presents a progressive two-stage training framework that decouples geometry-aware foreground transformation from background preservation and realistic video composition, without mesh-pixel alignment and explicit 3D reconstruction at inference.
Youze Huang, Penghui Ruan, Bojia Zi et al.
· 0 citations