Creating scientific visuals from text with AutoFigure toolkit

A tutorial has demonstrated how to use AutoFigure, a toolkit designed to generate scientific figures directly from text descriptions and structured explanations. The guide walks through setting up the AutoFigure environment, resolving dependency conflicts such as Pillow compatibility, and preparing tools for SVG and PNG outputs. It also covers building a custom reference figure, configuring an API-based workflow, and converting detailed system descriptions into publication-style diagrams. Offline rendering tests for SVG and PNG were conducted to ensure the pipeline functions correctly before any API calls. The tutorial includes steps for importing paths, defining provider settings, and preparing figure descriptions and sample paper content. Helper functions were created to run commands, print section headings, read files safely, and securely collect API keys. Utility functions for displaying generated files in Colab, including PNG, SVG, JSON, Markdown, and draw.io outputs, were also developed. An HTML gallery generator was built to review all outputs on a single page, and a result-summary function was added to display generation metadata and previews. System packages were installed, Pillow compatibility issues resolved, and the AutoFigure repository cloned. The SDK and its dependencies were installed, and the main classes and utilities were imported after confirming the environment was ready. A custom reference image was created to guide the layout style, and AutoFigure was configured with provider, model, API key, output directory, reference image, iteration settings, and visual style. The internal prompt template was previewed, and the main text-to-figure workflow was executed to produce a scientific diagram. A paper-style Markdown file was created, and AutoFigure’s methodology extractor was optionally used to generate a figure from paper content. A simple PDF version of the paper was also created, and the text extraction pipeline was tested. The mxGraph draw.io workflow was optionally run, all generated files were listed, an HTML gallery was built, and the complete output folder was exported as a zip archive. The tutorial concluded by showcasing a full AutoFigure workflow from environment setup to figure generation, validation, previewing, and export. It highlighted how AutoFigure transforms complex research or system descriptions into structured visuals while allowing control over references, style, output format, iterations, and optional paper-based extraction. The final result is a Colab-ready pipeline capable of generating SVG figures, preparing editable draw.io outputs, testing PDF extraction, and packaging all generated assets for future use.