Challenges and Design Opportunities for AI-Based Tutoring and Assessment Software in Special Education: An Interview Study with Teachers (EC-TEL '25)

Placebo Effect of Control Settings in Feeds Are Not Always Strong (CHI '25)

How Should We Design Technology With Diverse Stakeholders Who Wish Not to Attend Design Activities Together? (CHI '25)

The effects of physical coherence factors on presence in extended reality (IJHCS '23)

OmniQuery: Contextually Augmenting Captured Multimodal Memories to Enable Personal Question Answering (CHI '25)

"We happen to be different and different is not bad": Designing for Intersectional Fat-Positive Information-Seeking (CHI '24)

What About My Design Context?: Exploring the Use of Generative Al to Support Customization of Translational Research Artifacts (DIS '25)

Amuse: Human-Al Collaborative Songwriting with Multimodal Inspirations (CHI '25)

Code Shaping: Iterative Code Editing with Free-form AI-Interpreted Sketching (CHI '25)

GPT-4V Cannot Generate Radiology Reports Yet (NAACL Findings '25)

Challenges and Design Opportunities for AI-Based Tutoring and Assessment Software in Special Education: An Interview Study with Teachers (EC-TEL '25)

Placebo Effect of Control Settings in Feeds Are Not Always Strong (CHI '25)

How Should We Design Technology With Diverse Stakeholders Who Wish Not to Attend Design Activities Together? (CHI '25)

The effects of physical coherence factors on presence in extended reality (IJHCS '23)

OmniQuery: Contextually Augmenting Captured Multimodal Memories to Enable Personal Question Answering (CHI '25)

"We happen to be different and different is not bad": Designing for Intersectional Fat-Positive Information-Seeking (CHI '24)

What About My Design Context?: Exploring the Use of Generative Al to Support Customization of Translational Research Artifacts (DIS '25)

Amuse: Human-Al Collaborative Songwriting with Multimodal Inspirations (CHI '25)

Code Shaping: Iterative Code Editing with Free-form AI-Interpreted Sketching (CHI '25)

GPT-4V Cannot Generate Radiology Reports Yet (NAACL Findings '25)

Transform any paper into

short-form videos.

Harnessing generative AI models, our system automatically transforms a PDF paper into engaging video scripts and visual concepts.
You can further customize these videos by providing your own style and audience context.

1. Upload a PDF & select a script

Upload your academic paper. Our system automatically analyzes the content and generates several engaging "hook" and script options. Choose your favorite and refine the text and voiceover tone in the editor.

2. Generate video scenes in a storyboard

Based on your finalized script, generate corresponding video clips scene-by-scene. You can iteratively refine the script for each segment and regenerate visuals until you're satisfied with the entire storyboard.

3. Add credits & generate the final video

Finalize your video with an automatically generated credit screen that attributes the original authors and you as the creator. Then, merge all scenes into a final, shareable short-form video.

Publication

PaperTok: Exploring the Use of Generative AI for Creating Short-form Videos for Research Communication

Meziah Ruby Cristobal*, Hyeonjeong Byeon*, Alex Chen*, Ruoxi Shang*, Donghoon Shin*, Ruican Zhong*, Tony Zhou*, Gary Hsieh

*co-first authorship

CHI 2026 PDF Project website

FAQ

  • Academic papers, while rich in knowledge, can be dense and time-consuming to digest for a broad audience. Our system aims to bridge this gap by automatically transforming complex research papers into engaging, digestible short-form videos, making academic insights more accessible and shareable. It's unrealistic to expect researchers to do this themselves due to time constraints, so we built this interface to scale up the creation of video summaries using generative AI.

  • The system generates video scripts, key visual suggestions, and audio narration based on the paper's content. The goal is to produce concise, attention-grabbing summaries suitable for platforms like TikTok, Reels, or YouTube Shorts, highlighting core findings and implications.

  • While we've primarily tested and demonstrated efficacy with HCI papers, there are no strict requirements regarding the paper topic or discipline. Our system can process various research domains to extract key information for video summarization.

  • The generated video content, scripts, and any provided feedback will be saved in our system and may be used for future analysis, research, and improving the video generation model.

  • Yes! The analysis of our system will be published, and the codebase can be found here.

This work is supported by an from Google

Made with ♥ by