MirrorWorld: Taming Video Diffusion Models for Mirror Reflection Generation
The paper targets a specific failure in video generation: mirrors that do not match the scene around them.
MirrorWorld treats mirror reflection generation as video inpainting with explicit scene-to-mirror modeling. Its two added pieces split the task into what should appear in the reflection and how that content should be arranged. The authors also build a benchmark from four existing video mirror datasets. They report better reflection reconstruction than image-based reflection methods and strong video inpainting baselines. ArXiv · AI/CL/LG's note
MirrorWorld treats mirror reflection generation as video inpainting with explicit scene-to-mirror modeling. Its two added pieces split the task into what should appear in the reflection and how that content should be arranged. The authors also build a benchmark from four existing video mirror datasets. They report better reflection reconstruction than image-based reflection methods and strong video inpainting baselines. ArXiv · AI/CL/LG's note
score 4