Paper2Video - A demonstration video project for generating academic papers launched by a national university.

What is Paper2Video?

Paper2Video, developed by the Show Lab at the National University of Singapore, automatically generates presentation videos from academic papers. Using the PaperTalker multi-agent framework, it transforms papers into complete presentation videos including slides, subtitles, audio, and speaker avatars. The framework comprises four modules: a slide builder, a subtitle builder, a cursor builder, and a speaker builder, responsible for slide generation, subtitle generation, cursor positioning, and speaker video generation, respectively. Paper2Video provides the first high-quality academic presentation video benchmark, containing 101 papers and their corresponding author presentation videos, slides, and other data. The benchmark also includes four evaluation metrics: Meta Similarity, PresentArena, PresentQuiz, and IP Memory, used to measure whether the video accurately conveys the core ideas of the paper, is easy to understand, highlights the author's contributions, and enhances the research's impact.

Paper2Video's main functions

• Automatic video generationIt automatically generates presentation videos from academic papers, transforming complex academic content into easily understandable visual and auditory forms.
• Multi-agent frameworkBy integrating multiple modules such as slide generation, subtitle generation, cursor positioning, speech synthesis, and speaker avatar rendering through the PaperTalker framework, we can achieve efficient and high-quality video production.
• High-quality benchmark datasetsIt provides a benchmark containing data on 101 papers and their authors' presentation videos, slides, etc., to provide a standard for the research and evaluation of academic presentation videos.
• Customized evaluation indicatorsEvaluation metrics such as Meta Similarity, PresentArena, PresentQuiz, and IP Memory were designed to measure the quality and effectiveness of the demonstration videos from different perspectives.
• Easy-to-use toolsIt provides complete code and detailed usage guidelines, making it easy for researchers and developers to use the tool to generate their own demo videos.

Paper2Video's technical principles

• Slide generation and optimizationExtract content from the LaTeX source file of the paper to generate a Beamer format slide draft. Optimize the layout using a "tree search visual selection" method; the system generates multiple layout candidates, and then a Visual Language Model (VLM) evaluates the best version.
• Subtitle and cursor generationThis tool generates corresponding scripts (subtitles) for the slides and plans the trajectory of the mouse cursor as the speaker presents. The cursor movement and speech are precisely aligned in time and space to guide the audience's attention.
• Speaker generationUsing a portrait photo of the author and a short audio sample, a virtual avatar with the author's personal features and synchronized lip movements and speech is synthesized through text-to-speech (TTS) and speaker face generation technology.
• ParallelizationThe video generation task was split into slides and processed in parallel, which greatly shortened the total generation time.

Paper2Video's project address

• Project official websitehttps://showlab.github.io/Paper2Video/
• Github repositoryhttps://github.com/showlab/Paper2Video
• arXiv technical paper: https://arxiv.org/pdf/2510.05096

Application scenarios of Paper2Video

• Academic ConferenceIt provides researchers with tools to quickly generate high-quality presentation videos, saving preparation time and improving presentation effectiveness.
• Online coursesIt helps educators transform academic paper content into engaging video courses, enhancing the interactivity and appeal of their teaching.
• Social media disseminationThis enables academic findings to be shared on social media in a more accessible and understandable video format, thereby expanding the impact of research.
• Academic ReportIt facilitates researchers in quickly creating academic report videos for internal presentations or public lectures.
• Research and PromotionIt provides a new way for research institutions and scholars to showcase their research findings, thereby increasing the visibility and public awareness of their research.