Researchers at the University of Washington's Prosocial Computing Group have built PaperTok, an AI tool that turns research papers into 45-second videos. Presented on April 17 at the Association for Computing Machinery Conference on Human Factors in Computing Systems in Barcelona, the system uses Google Gemini to generate scripts and storyboards, aiming to put scientists in direct control of how their work appears on short-form video platforms.
The project started when students noticed a wave of non-scientists using generative AI to create science explainers, a trend that risked accelerating misinformation. "The alternative is that science is being talked about without scientists," said co-lead author Meziah Ruby Cristobal, a UW doctoral student in human centered design and engineering.
Senior author Gary Hsieh, a UW professor in human centered design and engineering, said that reading dense research papers remains a challenge even for academics operating outside their core field. "So we wanted to find a way to quickly turn papers into a format that laypeople would want to engage with, and we wanted to study how they engaged with it," he said.
How PaperTok works
Currently accessible only to users with a paid Google Gemini subscription, PaperTok accepts an uploaded research paper and immediately suggests four hook options to open the video. For a demo video about PaperTok itself, one hook reads: "Ever get overwhelmed reading a dense academic paper?" The team interviewed eight science communicators to land on the hook format. "We found that hooks are integral to short-form videos," said co-lead author Donghoon Shin, a UW doctoral student. "Because you're competing with other videos online, you have only a few seconds to grab someone's attention."
Once the researcher picks a hook, the tool's text-to-video pipeline generates a script. The author can edit the transcript iteratively, then move to a storyboarding phase where the script is divided into scenes. Users refine the script and accompanying video clips until satisfied, and can add a byline that appears alongside the paper's authors at the end.
How researchers responded
The team asked 100 online participants and 18 academic participants to compare PaperTok's output with videos from two other PDF-to-video generators. Respondents found PaperTok easy to use and rated its videos as more engaging. But some flagged a concern: the presence of AI artifacts, such as nonsense text, made a few videos feel "too AI-ish." Those users said they would hesitate to share the clips publicly, fearing it might hurt the credibility of their scholarship.
Why this matters for researchers
PaperTok signals a deliberate push to keep domain experts in the loop when generative AI crosses into science communication. Cristobal said the team's core question was, "How can we enable researchers to create engaging short-form videos?" She added: "With generative AI tools, anyone can generate a video from a PDF in minutes, and that presents all sorts of problems-misinformation, AI slop. So we wanted to build a tool that keeps humans, ideally experts, involved. If anything, we hope that PaperTok highlights how important people are in science communication."
For researchers exploring AI for Science & Research, PaperTok illustrates a model where AI accelerates production without removing the scientist from the editorial seat. The lab plans to deepen customization, including ways to let users draw directly on storyboard frames so the output aligns more closely with their intent.
Your membership also unlocks:
