# Scientific Figure capabilities > Turn text, sketches and PDFs into publication-ready scientific figures with AI. The limits and model facts below are the ones the application enforces, read from its configuration. Pricing and credit costs are in https://scientificfigure.com/pricing.md; how to credit the tool in a paper is in https://scientificfigure.com/cite. ## Inputs - A written description of the figure in plain language. No prompt syntax to learn. - A hand sketch or a reference image, used to set the layout or the look of the figure. - An abstract or a full paper PDF; the figure is drawn from the paper's own content. - Tabular data for chart tools (box plot, bar graph, histogram, heatmap, line graph, scatter plot, radar chart), where the geometry is computed from the numbers and the model only draws. - Typed parameters for computed figures (Bode plot transfer function, Lewis structure formula, probability tree outline, bell curve mean and SD), read by a short extraction step and computed in code before drawing. ## What it produces - Scientific illustrations - Graphical abstracts - Conference posters - Mechanism diagrams - Signaling pathway diagrams - PRISMA flow diagrams - Phylogenetic trees and cladograms - Cell diagrams - Circuit diagrams Each output type has a dedicated tool page under https://scientificfigure.com/tools, and the same tools are available inside one workspace. ## Capabilities - Generate figures from a text prompt, a hand sketch, an abstract or a full paper PDF - Every label, arrow, color and shape stays editable after generation — not a flat image - Export to PNG, JPG, layered SVG and editable PPTX - Up to 8K export, 600+ DPI at single-column width - Recolor to journal palettes including Nature, Science, NEJM, The Lancet, JAMA and Okabe-Ito - Nine illustration styles, from clean schematic to 3D render - Conference poster layouts at A0, A1 and 48×36 inches - Built-in vector canvas for editing without leaving the browser - Publication and commercial rights to everything you generate, on any paid plan or credit pack - Private uploads are encrypted and never used to train models ## Image models | Model | Highest output | Takes an input image | On the free plan | |---|---|---|---| | FLUX 2 Pro | 4K | No | No | | GPT Image 2 | 4K | Yes | Yes | | GPT Image 2.5 Flare | 4K | Yes | Yes | | GPT Image 2.5 Sunburst | 4K | Yes | No | | Nano Banana | 1K | Yes | No | | Nano Banana 2 | 4K | Yes | No | | Nano Banana Pro | 4K | Yes | No | | Seedream 5.0 Pro | 2K | No | No | The default model is GPT Image 2. A model that takes an input image can edit an existing figure and work from a sketch or reference; the others generate from text only. ## Resolution - 1K: 1,536 px on the long side (the free plan renders at this tier only) - 2K: 2,048 px on the long side - 4K: 4,096 px on the long side - Upscale after rendering: 2K = 2,048 px, 4K = 4,096 px, 8K = 8,192 px. Upscaling is a super-resolution pass, not a re-render, so labels and data do not change. 8K is reserved for Pro, Plus and credit-pack accounts. - Exported PNG and JPG files carry a DPI tag matched to their pixel width: 1K is enough for a single-column figure at 300 dpi, 2K and above for double-column and 600 dpi. ## Editing A generated figure is not the end of the conversation. In the workspace the figure can be changed by typing what to change, or with one-click actions: - Quick edit: describe a change in words and the figure is redrawn with it - Redraw a region: brush over the part to change, describe the replacement - Change text: correct a label or caption without touching the rest - Upscale: pixel-exact enlargement to 2K, 4K or 8K, no re-render - White background: remove a tinted or textured background - Recolor: apply a journal or custom palette to the whole figure Edit in Canvas goes further: it decomposes a finished figure into editable objects (text blocks, panels, shapes, traced vector paths) on a browser canvas, where each can be moved, restyled or rewritten, and the result exports as layered SVG or editable PPTX. It costs 150 credits per figure, priced in credits rather than tied to a plan. ## Export - Formats: PNG, JPG, SVG, PPTX. - SVG exports are real vectors: text stays text, shapes stay shapes, and the file opens for editing in Illustrator, Inkscape or Figma. - PPTX exports are editable slides with native text and shapes, not a pasted image. - Free-plan exports carry a small corner watermark; paid plans and packs export clean files. ## Styles and palettes - Nine illustration styles, from clean flat schematic to 3D render, chosen per figure. - Journal palettes: Nature, Science, NEJM, Lancet, JAMA. Colorblind-safe palettes: Okabe–Ito, Viridis, Mono. 23 presets in all, plus custom palettes of up to four colors. - Any palette can be applied before generation or to a finished figure with the recolor action. - Conference poster layouts at A0, A1 and 48 × 36 inches. ## Rights, licensing and privacy - Everything generated on a paid plan or with pack credits may be published and used commercially. - Every export on a paid plan or pack is issued a numbered publication license certificate, which anyone can confirm at https://scientificfigure.com/verify. Details: https://scientificfigure.com/publication-license - Uploads are private, encrypted and never used to train models. - Suggested figure credits, methods statements and an AI-disclosure line: https://scientificfigure.com/cite ## Who it is for Researchers, Graduate students, Principal investigators, Postdoctoral researchers, Educators, R&D teams, Medical writers, Science communicators. ## How it differs - Unlike template libraries such as BioRender, figures are generated rather than assembled from a fixed icon set, so the output is not limited to shapes someone already drew. - Unlike general-purpose AI image generators, output follows scientific drawing conventions and stays editable, so a wrong label can be corrected instead of regenerated. - Unlike vector editors such as Illustrator or Inkscape, no design skill is required to get a first draft, and the result still exports as real vectors those tools can open. ## Getting started 100 free credits on signup, no card required. Sign in at https://scientificfigure.com/sign-in, describe the figure, and the first draft is usually ready in under a minute.