Tired of reading academic papers? This AI tool transforms them into videos similar to TikTok.

      PaperTok makes academic papers accessible for those with short attention spans

      Academic papers may not be seen as enjoyable reading material. For the average person, the lengthy, dense, and specialized language can be quite daunting. Typically, these papers are presented as PDFs that seem designed to challenge one’s resolve. A team of researchers from the University of Washington believes there is a better method to present this work to the general public. They are transforming academic papers into brief videos.

      This tool is named PaperTok and it employs AI to assist researchers in converting academic papers into concise, social media-friendly videos. Picture TikTok, but focused on explaining research rather than disseminating mind-numbing content. The most appealing aspect? PaperTok is creating a platform that involves scientists and researchers in the process rather than solely using AI to summarize the papers.

      How PaperTok operates

      A researcher uploads their paper to PaperTok, and the platform utilizes Google Gemini to analyze the document and propose brief video concepts. The system offers various hooks, as engagement in short-form videos hinges on the initial seconds. After selecting a hook, PaperTok generates a script. The researcher can then modify the script, alter the tone, and refine the content prior to progressing to the storyboard phase. The platform subsequently divides the script into scenes and produces visual clips section by section.

      After the video is completed, PaperTok ensures to credit the authors of the paper and the video creator, affirming that AI does not overshadow those who contributed the actual work. The videos are kept short, with PaperTok capable of transforming research papers into 45-second videos, aligning much closer with how many people consume media online.

      Using AI to combat AI

      PaperTok was developed because researchers observed that individuals were already employing generative AI to produce brief science videos. However, non-experts using AI to interpret complex papers often result in errors that propagate misinformation rapidly. Therefore, PaperTok applies these tools correctly, while keeping scientists involved in the process.

      In tests, UW researchers evaluated PaperTok videos against those from two other PDF-to-video converters using feedback from 100 online participants and 18 academic participants. The PaperTok videos were regarded as more user-friendly and engaging. Nonetheless, some users still felt the videos appeared excessively AI-generated, exhibiting issues such as odd visual artifacts or nonsensical text, indicating that the technology is not yet flawless. A video that seems overtly AI-made might even damage its credibility.

      Currently, PaperTok is available only to users with a paid Google Gemini subscription. The team also aims to enhance customization options, allowing users to provide more specific guidance for certain aspects of the generated scenes.