Shape-Aware Text-Driven Layered Video Editing

Figure 1 . Shape-aware consistent video editing. Our method enables consistent text-guided video editing with both appearance and shape changes. The top row shows the input frames. The second and third rows present editing results from two text prompts: "running sports car" and "running minivan", respectively. Note that text-driven editing involves both texture and structure editing on the foreground object. Our method performs consistent edits on sequential frames while preserving the object motion in the input video.