FigureWeave v0.1.0

FigureWeave is an open-source workflow for turning method text into editable scientific figures.

I built it to make paper figure creation less manual: starting from a method description, FigureWeave generates a draft image, segments visual elements locally with GPU-accelerated SAM3, and reconstructs the result into an editable SVG that can be refined in the browser.

This first public release includes:

• the FigureWeave codebase and local workflow
• multi-candidate generation
• dual-provider routing for image drafting and SVG reasoning
• browser-based editable SVG output
• local CUDA-accelerated post-processing
• showcase examples and UI previews

It is designed for workflows such as:

• method overview figures
• system pipeline diagrams
• model architecture sketches
• editable draft figures for papers and slides

Full Changelog: https://github.com/Krisocer/FigureWeave/commits/v0.1.0