AI Inspiration WorldWhat it is
AutoFigure is a scientific-figure tool by the AutoFigure Team (GitHub ResearAI); this collection lists the original together with its follow-up, AutoFigure-Edit. The original uses large language models (LLMs) to read your text or paper, pull out the method, draw a figure, then review and revise it over several rounds, exporting SVG or mxGraph XML. The authors call AutoFigure-Edit the next version: it segments the elements of a reference or generated image and rebuilds them as editable SVG, with a built-in editor. You still need to check the science in the figure yourself.
Who it's for:Researchers who need method diagrams for papers and want to keep editing them later. Self-hosting takes some Python.
Before you start: The original needs Python 3.8 or later, installs with pip -e ., and renders with Playwright's Chromium. AutoFigure-Edit needs Python 3.10 or later and either a separate SAM3 segmentation model or a Docker deployment. You bring your own model and segmentation accounts, which may cost money; the authors say the AutoFigure-Edit online platform is free for scholars. It reads the paper PDF or Markdown, descriptions and reference images you give it, and writes outputs, intermediate files and logs locally. The model providers you pick (such as OpenRouter, Google or OpenAI) receive the text and images used for generating and reviewing; AutoFigure-Edit uses Roboflow for segmentation by default, so your images go there. If you use these features, weigh the safety risks yourself. AI Inspiration World is not responsible for any consequences. Both repos are MIT-licensed, and the name and logo have separate trademark terms.
Based on the project's public documentation. Not tested by AI Inspiration World.