Google's latest tool is a game-changer—automating illustration creation for academic papers! PaperBanana (later renamed PaperVizAgent) is an AI-powered tool designed to automatically generate illustrations for academic papers. Here’s what it does: - Converts the methodology text in your paper into publication-quality academic diagrams (like flowcharts, architecture diagrams, method diagrams, etc.). - Generates statistical plots as well. - Uses multi-agent collaboration to complete the process: retrieve reference images → plan content → adjust academic style → generate illustrations → self-critique and iterative optimization. In simple terms, it helps researchers spend less time drawing and formatting, turning “text-based explanations” into visually appealing, standardized academic illustrations. The authors of the paper are from Peking University and Google Cloud AI Research, aiming to lower the barrier for creating paper illustrations. The tool is open-source and requires configuring Gemini or other model API keys to use. Paper: https://arxiv.org/abs/2601.23265v1 GitHub repositories: Author-maintained repository: https://(((github.com)))/dwzhu-pku/PaperBanana This version is maintained by the first author, Dawei Zhu, forked from Google Research’s official repository and continuously updated (supports Gradio/Streamlit, Hugging Face Spaces demos, OpenRouter, etc.). The README clearly states that the original version was open-sourced by Google Research. Official original repository (Google Research): https://(((github.com)))/google-research/papervizagent The initial open-source version of the paper, later renamed PaperVizAgent (formerly PaperBanana). Project homepage: https://dwzhu-pku.(github.io)/PaperBanana/ Online demo: https://huggingface.co/spaces/dwzhu/PaperBanana There are also some community reproductions (e.g., https://(((github.com)))/llmsresearch/paperbanana).