researai/autofigure-edit

AutoFigure-Edit is a Python tool that converts the method section text from a scientific research paper into a fully editable vector diagram, specifically an SVG file. SVG stands for Scalable Vector Graphics, a format where every shape, label, and line is defined as editable code rather than a fixed pixel image. The problem it solves is the tedious work of manually drawing figures for academic papers. Researchers typically spend significant time in illustration software recreating diagrams that describe how their method works. AutoFigure-Edit takes the written description and generates a draft diagram automatically, then outputs it in an editable format for refinement. The pipeline runs in four stages. First, a language model generates a raster draft image from the method text. Second, a segmentation model called SAM3 detects the distinct icon and text regions in that draft. Third, the system constructs a structural SVG wireframe using consistent placeholders. Fourth, the final SVG is assembled with high-quality icon crops mapped to those placeholders. The result opens in an embedded browser-based editor where you can modify text, shapes, and layout directly. Additional capabilities noted in the README include style transfer, the system can mimic the visual style of a reference image you supply, and the ability to skip the generation step if you already have an existing figure you want to vectorize. You would use this if you are a researcher who needs publication-quality method diagrams and wants to start from an AI-generated draft rather than a blank canvas. The tech stack is Python. The full README is longer than what was provided.