University of Washington Develops AI Tool to Convert Research Papers into Videos

By Priya Anand

The University of Washington's Prosocial Computing Group has launched PaperTok, an AI-driven system designed to convert academic research papers into short, engaging videos. This initiative, led by Professor Gary Hsieh, addresses the growing trend of non-experts using generative AI to produce science videos, often lacking scientific accuracy.

PaperTok's workflow involves generating 45-second videos that start with an attention-grabbing hook, followed by AI-generated scripts and segmented scene-based storyboarding. The system includes an editable interface for refining video clips, ensuring the final product attributes both the original paper's authors and the human creator.

The system was evaluated through a study involving 18 researcher participants and 100 crowd-sourced audience members. Results showed that videos produced by PaperTok were rated higher in engagement and entertainment compared to those from existing platforms like SciSpace and PDFtoBrainrot, while maintaining similar levels of informativeness.

Co-lead author Meziah Ruby Cristobal highlighted the importance of scientists being involved in the communication of their work, stating, “The alternative is that science is being talked about without scientists.” Fellow co-lead author Donghoon Shin emphasized the role of hooks in short-form videos, noting the critical need to capture audience attention quickly.