PaperFig
AI scientific figure generator built for researchers
About PaperFig
PaperFig is a web-based AI scientific figure generator aimed at researchers who need mechanism diagrams, pathway illustrations, experimental workflows and graphical abstracts for papers, slides and lab meetings. Instead of arranging icons, arrows and text boxes by hand in a drawing program, you describe the figure in plain language, or hand it a rough sketch, a lab photo, a PDF or an existing figure, and PaperFig produces an illustrated draft you can then correct and export. The site positions it as a way to get from an early idea to an editable draft faster, for example when a reviewer asks for one more pathway panel shortly before a resubmission deadline.
The tool is organised around research disciplines rather than generic image styles. Its public gallery holds 84 curated example figures across seven fields - biomedical, materials, environmental, energy, bioengineering, CS/AI and geoscience - each with the prompt that produced it, so you can copy a prompt close to your own topic and adapt it. PaperFig is open about the limits of AI output: its FAQ says generated labels can be wrong, keeps label correction inside the workflow, and states that the researcher remains the final scientific reviewer. According to its terms, generation currently runs on Google Gemini image model families, and the service does not claim ownership of figures you create. Pricing is credit-based, with a free signup allowance and paid monthly or yearly plans that add publication and commercial rights.
You start on the Create page and choose an input path: Text to Figure, Sketch to Figure, Photo to Figure, Reference to Figure, or the Figure Enhancer for improving a draft you already have. PDFs can also be used as source material. For a text prompt, you describe the mechanism, components, relationships and labels you want shown, for instance a hydrogel sensor with its conductive network, skin-mounted patch and signal output.
Next you pick a generation mode. Fig Standard is the faster option for drafts and everyday figures, while Fig Refined renders in higher detail for final polish and may take longer; both cost about 50 credits per figure. You can leave the aspect ratio on Auto or set 1:1, 4:3, 16:9 or 3:4 to fit a journal column or a slide.
Once a draft appears, you check it. The built-in OCR text editor detects the labels on the image so you can click a single word and fix it without regenerating the whole figure. When the figure is right, you download a PNG, optionally upgrade it to a Faithful 4K PNG for submission, or, after running Fix Labels, export editable SVG or PPTX files to keep working in Illustrator, Inkscape or PowerPoint. Paid plans add TIFF and PDF export and remove the watermark.
- •Five input paths - Start from a written description, a hand-drawn or whiteboard sketch, a photo of a device or experiment, a reference figure whose structure you want to borrow, or a PDF. The Figure Enhancer takes an existing draft and works on labels, spacing, hierarchy and arrows for clearer reading.
- •Standard and Refined generation modes - Fig Standard produces quicker drafts for everyday work, and Fig Refined spends more time on detail for figures headed to a manuscript. Aspect-ratio presets help match the output to the place it will be used.
- •OCR label editing - Labels on the generated image are detected as text, so a misspelled term or wrong gene name can be corrected in place. This directly addresses the most common weakness of AI-generated scientific images, which is unreliable lettering.
- •Editable and high-resolution exports - Core downloads are PNG, with a Faithful 4K PNG upgrade for final images. After label fixes, figures can be exported as editable SVG or PPTX, and paid plans add TIFF and PDF output for journal submission systems.
- •Discipline gallery with reusable prompts - The inspiration library groups real journal-style examples by field and by mode, and each example exposes its prompt so you can copy it as a starting point rather than writing from a blank page.
- •Research preparation tools - Alongside the generator, PaperFig offers a Graphical Abstract Maker for planning a single visual message, a Scientific Figure Prompt Generator that structures a topic into panels, labels and relationships, a Journal Figure Checker covering resolution, dimensions, color, typography and file format, and a free figure drawing tool that works without logging in.
PaperFig is built for academic researchers, PhD students and postdocs who need schematic figures but do not have the time or illustration skills to build them by hand. Typical uses include mechanism and pathway figures for biomedical papers, material structure and device diagrams in materials science and energy research, environmental and geoscience process figures, and method workflows for computational biology and AI papers.
Graduate students preparing a thesis or a first manuscript can use the free allowance to test whether the drafts suit their field before paying. Active researchers facing revision rounds or grant deadlines are the target of the mid-tier plans, which the pricing page frames around the number of figures produced per month. Research groups can use the Lab Pro plan for lab-wide iteration, and the site notes that payment records are available for grant or institutional reimbursement.
It also fits teaching: the same mechanism prompt can be turned into a cleaner explanatory figure for a course or lab training session. Anyone planning to publish should note the licensing split - free downloads are watermarked and limited to non-commercial, non-publication use, while journal publication, preprints, posters, grant documents and commercial use require a paid plan or a credit pack purchase. Users should also follow the guidance not to upload patient data, unpublished confidential results or restricted institutional material.
Pricing
$0 - $72/mo
- Free$0/mo
- Scholar$16/mo
- Pro Scholar$27/mo
- Lab Pro$72/mo
- Mini Pack$9 once
- Light Pack$45 once
- Standard Pack$90 once
- Heavy Pack$180 once
- Mega Pack$270 once
From the vendor pricing page, 2026-10-29
