UW Researchers Use Gen AI to Create TikTok Science Videos

Researchers at the University of Washington’s Prosocial Computing Group have developed PaperTok, a generative AI tool designed to convert dense academic research papers into engaging, short-form TikTok videos. This initiative aims to make scientific findings more accessible to the general public, addressing the common challenge of people not reading lengthy research articles. The tool utilizes Google's Gemini AI to analyze uploaded papers and generate concise scripts, which are then transformed into video clips.

PaperTok was created in response to observing non-scientists using generative AI to produce science content on social media, which raised concerns about accuracy. The University of Washington team sought to provide a scientifically grounded method for summarizing research in a format that resonates with a broader audience. Senior author Gary Hsieh, a professor in human-centered design and engineering at UW, stated that the goal was to find a way to quickly adapt papers into a format laypeople would find engaging and to study their interaction with this content. The research was presented at the Association for Computing Machinery Conference on Human Factors in Computing Systems in Barcelona in April.

The PaperTok platform allows users to upload a PDF of a research paper. The system then analyzes the content and offers four potential "hooks" or opening lines for the video. Users select their preferred hook, and PaperTok generates a script that can be further edited for clarity and tone. Following script finalization, the tool generates corresponding video scenes, presented in a storyboard format, which users can also refine. Currently, access to PaperTok requires a paid subscription to Google Gemini. The generated videos are approximately 45 seconds long, suitable for platforms like TikTok.