Paper2Video: Automatic Video Generation from Scientific Papers
Primary Source: https://showlab.github.io/Paper2Video/
Author: Show Lab (NUS) · Year: 2026 · Publisher: Show Lab
Summary
An automated pipeline that parses LaTeX/PDF scientific papers, extracts core figures and equations, generates an engaging explanatory narrative, and synthesizes full multi-scene presentation videos.
Key Points
- Two-stage framework: content abstraction with slide layout planning, followed by multimodal video synthesis.
- Automated slide animation, synthetic voiceover narration synchronization, and visual focus highlights.
- Standardized benchmark dataset and evaluation metrics for automated academic presentation quality.