Sign up now — your account and credits will be ready when you switch to a larger screen.
PDF-TO-FIGURE
Turn your paper into figures, page by page.
Drop a PDF — SciFig reads it, extracts key concepts, and generates publication-ready figures matched to your paper's content.
AI will extract concepts from your PDF
How it works
PDF to scientific figure, in 3 steps
1
Drop in your paper
Methods, results, or review — PDF up to 30MB.
2
AI extracts concepts
Core findings, mechanisms, and pathways detected automatically.
3
Pick what to visualize
Choose a concept for SciFig PDF to figure to render. Edit text anytime.
PDF-to-Figure FAQ
Common questions about generating figures from PDF documents.
1.
Yes. Your unpublished draft stays yours, full stop.
Zero AI Training: Uploaded PDFs are read only to build your generation prompt. Never recycled into model training — ever.
Draft-Safe by Default: Pre-publication manuscripts, confidential grant drafts, embargoed data — all treated identically to any other upload. No exceptions.
Yours to Delete: Remove any uploaded PDF from our servers at any time. Zero retention if you don't want it.
2.
Drop the manuscript. SciFig PDF to figure reads it. You get a graphic abstract.
Manuscript In, Visual Out: Don't stare at a blank canvas after 40 pages of writing. Upload your draft PDF — the LLM reads the paper and drafts a generation prompt, producing a single-panel graphic abstract in minutes.
Your Angle, Your Prompt: The manuscript sets the science; your prompt sets the emphasis. Add a line like "focus on the core mechanism, not the assay controls" and regenerate until the angle is yours.
Journal-Ready on Demand: Output is high-res raster by default. One-click vectorize to SVG/PPTX when your target journal asks for editable layers.
3.
Your abstract is enough. A poster comes out of it.
Abstract-First Workflow: Paste your abstract as a PDF and SciFig PDF to figure reads the text, drafting the poster's visual scaffolding around the headline finding.
No Printing-Night Panic: Generate a draft poster in minutes instead of the night-before-the-conference scramble. Regenerate any section without rebuilding the whole layout.
Polish Where It Counts: Vectorize to PPTX when you need to hand-tweak titles, affiliations, and QR codes in PowerPoint before the print-shop cutoff.
4.
Point SciFig at Methods only. Get just the diagram.
Section-Selective Focus: You don't have to feed the whole paper. Upload your PDF and tell SciFig to read just the Methods section — intro, results, and discussion are ignored when the prompt is drafted.
Order Preserved: A linear workflow, a branching protocol, a multi-plate assay — whatever your Methods describes, the generated diagram keeps the step order intact.
5.
Grant draft in, roadmap out. Deadline intact.
Proposal-Aware Reading: Upload your proposal draft or LOI. SciFig reads Specific Aims, Research Plan, and timeline language, then drafts a visual roadmap showing aims, dependencies, and milestones.
Your Terminology, Same Figure: Call it a "research roadmap," "technical roadmap," or "Specific Aims figure" — the generated diagram covers the same job. Tune axes, swim-lanes, or phase colors with an added prompt.
Revise Per Reviewer Pass: Reviewer wants a clearer Aim 3? Regenerate just that panel instead of the whole figure. Keeps your resubmission deadline intact.
6.
Any paper in your reading list becomes a talk-ready infographic.
Reading-List Friendly: Upload a paper you're reviewing or referencing — the LLM distills the core finding into a plain-language prompt, then draws a new infographic from that prompt.
Talk, Classroom, Science-Comm: Same flow for conference talks, lab meetings, teaching slides, and popular-science articles. The output is a fresh visual built from the summary — no figure lifting from the source.
One-Step Attribution: Put "include an attribution line 'Adapted from [Author, Year]'" into your prompt and SciFig renders the citation directly in the figure — no post-generation canvas edit needed.
7.
One PDF, many scientific figures. SciFig PDF to figure runs one section at a time.
Section-Selective Focus: Tell SciFig which section to read — Methods, Results, Discussion, Specific Aims, even a single paragraph. The LLM ignores the rest when drafting the prompt.
Full Figure Set From One Upload: Run the same PDF with different section targets to populate the lineup — graphic abstract from the abstract, method diagram from Methods, workflow from Results, roadmap from Aims.
8.
Upload supplies the science. Your prompt supplies the angle.
PDF Is the Content: Whatever's in the paper — mechanisms, data, methods — is what the LLM reads when drafting the generation prompt. You don't retype the science.
Your Prompt Is the Steering: Add a single line to focus the output: "emphasize the kinase cascade, not the downstream effects," "two-column layout, editorial palette," "omit the statistics panel." Without steering, you get SciFig's default interpretation.
9.
One PDF per generation — that's the current limit.
Single-File Focus: One PDF at a time keeps the LLM tightly focused on a single document's content.
Run It Again: Upload a different PDF on each generation — all results stay grouped in one project for easy comparison.
Synthesizing Multiple Sources: Need one figure that pulls from several papers? Summarize the key points yourself first, then use Text-to-Figure to generate from the summary.
10.
20 MB per file. No page cap. No OCR setup on your end.
Up to 20 MB per PDF: Covers almost every journal article, including figure-heavy papers. (Note: For 100 MB+ textbooks, split by chapter first.)
No Page Limit: Long reviews, theses, textbook chapters — the LLM reads end-to-end instead of truncating at page N.
OCR-Adaptive: Scanned PDFs from the 1990s and digital-native exports both work. You don't have to run OCR yourself — the LLM reads whatever is parseable.
11.
SciFig PDF to figure exports raster for most journals and vector for element-level edits.
Raster Is the Default Output: PNG/JPG up to 8K — clears 300 DPI at standard journal column widths. Enough for the large majority of submissions, talks, and posters.
Vectorize for Layered Editing: One click converts the figure to SVG/PPTX. Every shape becomes a separately selectable object in Illustrator, Inkscape, Figma, or PowerPoint.
12.
Three paths. Pick whichever is fastest for your fix.
Inline Text Edit: Click the label and retype it — no regeneration needed for typos, wrong gene names, or mislabeled proteins.
Reprompt the Section: For a wrong structure or a missing element, tighten the prompt ("the kinase is JAK2, not JAK1") and regenerate. Fastest for anything the LLM misinterpreted from the source.
Vectorize for Surgical Edits: Need to move an arrow, recolor one panel, or swap one icon? Vectorize first, then edit the specific shape — everything else stays intact.
