Helping Figures Tell their Story! Paper-Grounded Video Generation Explaining Complex Scientific Figures

In three linesNew arXiv paper introducing MINARD, a video generation system that transforms scientific figures into narrated walkthrough videos with region grounding. The pipeline generates paper-grounded narrations and sequentially aligns them to figure regions. Includes FigTalk benchmark with component-level grounding metrics.

Summary generated by Claude — human-verified