PaperClaw

PaperClaw is a web-based AI figure maker for researchers. It generates, edits, polishes and vectorizes mechanism diagrams, graphical abstracts, posters and statistical figures, with PNG and editable SVG export, on a credit-based plan.

PaperClaw is a web workspace built for one job: producing scientific figures. Instead of general AI art, it targets the visuals researchers need for manuscripts and teaching, such as mechanism diagrams, methodology flowcharts, statistical plots, graphical abstracts, educational infographics, academic posters and multi-panel layouts. A user types a plain-language description, picks a figure type, and receives a journal-styled draft in minutes. The workspace is organized around six tasks. Generate Figure turns text into an illustration. Edit Image changes an uploaded figure by instruction while leaving the rest untouched. Enhance removes backgrounds and upscales for print or slides. Sketch to Figure redraws a hand-drawn sketch or whiteboard photo while keeping its layout. Figure Polish restyles a rough draft with cleaner color, type and spacing without altering the data. Editable SVG creates true vector output from text or vectorizes an existing image, so the result can be refined in Illustrator, Inkscape or Figma. Raster export goes up to 4K PNG. Usage runs on credits. Each generation shown on the homepage costs 5 credits, and new accounts receive 10 free credits without a credit card. Paid plans are sold as monthly or yearly subscriptions, with options to buy credits and a team tier. Higher tiers add paper upload for automatic style retrieval, reference image upload and aesthetic refinement. The target audience is researchers, graduate students and educators who lack design skills or budget for a commissioned illustrator. Compared with BioRender, it has no fixed icon library, so figures start from the user's own idea, data or sketch. Compared with general image generators, it claims cleaner labels, axis text and notation. The site is candid that output can contain mistakes, so every label, value and symbol still needs a manual check before submission. It fits best as a fast first-draft engine for figures, followed by touch-ups in a vector editor.