Abstract

We introduce GroupSketch, a novel pipeline for vector sketch animation that effectively handles multi-object interactions and complex motions. Existing approaches struggle with these scenarios, either being limited to single objects or suffering from temporal inconsistency and generalization issues. Our method addresses these limitations through a two-stage approach: Motion Initialization and Motion Refinement. The first stage allows users to divide sketches into semantic groups and define key frames, generating a coarse animation through interpolation. The second stage employs our Group-based Displacement Network (GDN) to refine this animation by predicting group-specific displacement fields, leveraging priors from a text-to-video model. GDN incorporates specialized components including Context-conditioned Feature Enhancement (CCFE) to improve temporal consistency. Extensive experimental results demonstrate that our approach significantly outperforms existing methods in generating high-quality, temporally consistent animations for complex multi-object sketches, expanding the practical applications of sketch animation.