Researchers at the University of Washington have developed PaperTok, an AI-assisted system that helps turn academic papers into short, engaging videos for platforms such as TikTok. The project grew from a concern that generative AI was already being used to make science videos by people without scientific expertise, raising the risk of polished but inaccurate content. PaperTok aims to give researchers a practical way to participate in short-form science communication without removing human judgment from the process, tells Tech Xplore.
The system allows a researcher to upload a paper and uses Google Gemini to generate a 45-second video script. It first offers several possible hooks, since the opening seconds are critical for holding viewers’ attention. Users can then edit the script, refine the storyboard, adjust video clips, and add a byline that credits both the presenter and the paper’s authors.
The research team designed PaperTok after interviewing science communicators and content producers about what makes short videos engaging and credible. The tool was presented at the 2026 CHI Conference on Human Factors in Computing Systems in Barcelona.
In testing, 100 online participants and 18 academic participants compared PaperTok videos with videos from two other PDF-to-video tools. They found PaperTok easier to use and more engaging. Still, some users said the output looked too AI-generated, citing issues such as visual artifacts and nonsense text. They worried that such signals could weaken the credibility of their scholarship.
PaperTok’s larger value lies in its human-in-the-loop approach. The researchers plan to add more customization features so users can control specific visual elements. The work suggests that AI can help researchers reach wider audiences, but expert oversight remains essential for trustworthy science communication.