Track4Gen: Teaching Video Diffusion Models to Track Points Improves Video Generation
This addresses visual coherence issues in video generation for applications like content creation, though it is incremental as it builds on existing architectures like Stable Video Diffusion.
The paper tackles appearance drift in video generation by proposing Track4Gen, which integrates point tracking into video diffusion models to provide spatial supervision, resulting in reduced drift and improved temporal stability.
While recent foundational video generators produce visually rich output, they still struggle with appearance drift, where objects gradually degrade or change inconsistently across frames, breaking visual coherence. We hypothesize that this is because there is no explicit supervision in terms of spatial tracking at the feature level. We propose Track4Gen, a spatially aware video generator that combines video diffusion loss with point tracking across frames, providing enhanced spatial supervision on the diffusion features. Track4Gen merges the video generation and point tracking tasks into a single network by making minimal changes to existing video generation architectures. Using Stable Video Diffusion as a backbone, Track4Gen demonstrates that it is possible to unify video generation and point tracking, which are typically handled as separate tasks. Our extensive evaluations show that Track4Gen effectively reduces appearance drift, resulting in temporally stable and visually coherent video generation. Project page: hyeonho99.github.io/track4gen