New AI Tool Turns Academic Papers Into Accessible Short Science Videos for General Audiences

Researchers from the University of Washington’s (UW) Prosocial Computing Group spotted a clear growing trend across social media platforms: everyday members of the public were increasingly using generative artificial intelligence to create short, shareable science videos. The major downside, however, was that most of these amateur creators are not trained scientists — this pushes the already high risk of factual errors in AI-generated content even higher.

Intrigued by this unregulated creator trend, the UW research team built PaperTok, a custom platform that takes an uploaded full scientific research paper and uses Google’s Gemini model to generate a short, approachable script, then automatically converts that script into a ready-to-share video clip. As of now, PaperTok can only be accessed by users with an active paid Google Gemini subscription.

“Most people never read academic research papers for a wide range of reasons,” said Gary Hsieh, senior author of the project and a UW professor in human-centered design and engineering, in an official press release. “Even I struggle to parse papers in research areas outside my own expertise. That’s why we wanted to build a way to quickly turn full research papers into a format that non-experts actually want to engage with, while also studying how general audiences interact with this kind of adapted science content.”

The UW team presented their work on PaperTok this past April at the Association for Computing Machinery Conference on Human Factors in Computing Systems, held in Barcelona.