Any figure from a paper, book, or poster — style only.
2
Describe your subject
Tell AI what YOUR figure should show — not the reference's.
3
AI creates, never copies
SciFig reference to figure delivers a new scientific figure in the reference's style — safe to publish.
Reference-to-Figure FAQ
Common questions about generating figures from reference images.
1.
Upload a reference. SciFig reference to figure inherits its design DNA and applies it to your own content.
Style, Not Subject: We don't copy the reference's content. We learn its palette, line weights, and typography — then apply them to your own original figure.
Two Ways to Use It: Match a target-journal style you can't design yourself, or lock in one consistent look across every figure in your paper, slides, and poster.
Editable Output: The result is a clean figure in the borrowed style, with full editing available for your labels, arrows, and molecules.
2.
Don't design. Reference.
Borrow a Proven Style: Upload your target journal's figure template, a style guide you've been given, or your own best past figure. SciFig reference to figure extracts its palette, line weights, and typography for you.
No Color-Theory Guessing: Stop cycling through Adobe color wheels. The reference's colors are already pre-harmonized by someone who solved the problem before you.
Applied in One Click: No Illustrator, no "learn design." Pick a reference, hit apply — every panel inherits the look automatically.
3.
Upload one reference. Every new figure inherits the same DNA — automatically.
Series-Level Consistency: Your 10-panel figure, your methods schematic, your graphical abstract, and your conference poster all share the same palette, stroke weights, and typography. No more "these figures don't feel like they belong to the same paper."
One Reference, Every Output: Upload your first finished figure as the reference, and every subsequent figure auto-matches. No per-figure setup, no manual color-picking.
Cross-Format, Not Just Paper: The same reference applies to slide decks, posters, and grant figures — one look for your whole project, not five.
4.
Yes. Pick a preset or upload your own mood board.
Curated Presets: Morandi, editorial, clinical, monochrome, pastel, neon — one click applies a publication-grade palette.
Upload Your Own Mood Board: Paste Pinterest screenshots, textbook spreads, or your lab's branding. SciFig extracts the palette automatically.
Applied Consistently: The chosen palette carries across every panel in your figure, so your paper looks unified instead of random.
5.
Yes. Every element of your reference to figure output stays editable after the style transfer.
Click to Correct: Color looks slightly off? Palette too saturated for print? Click any element on the canvas and tweak — individual colors, line weights, or typography.
Re-Apply with a Different Reference: Don't love the first result? Swap in a different reference and re-apply — keep your content, change the aesthetic.
6.
Snap it. We clean it up.
Use Sketch-to-Figure: For hand-drawn diagrams, we have a dedicated mode — see `/sketch-to-figure` for the full workflow.
Photograph at Any Angle: Phone photo is fine. SciFig corrects perspective, removes paper texture, and redraws as clean vectors.
Handwriting Becomes Typography: Messy labels become clean journal-grade text automatically.
7.
Every format your reference-to-figure output needs for paper, deck, or poster.
Direct Exports: PNG (recommended for journal submission), JPG (smaller files for sharing or email), and PPTX (a 16:9 PowerPoint slide with the figure embedded — drop straight into a deck).
Vectorize for Layered Editing: Open the result in Vector Canvas and run Vectorize. One pass produces clean SVG (opens in Illustrator, Inkscape, Figma, or Affinity) plus an editable-shape PPTX where every element is a separately selectable PowerPoint object.
Resolution: PNG/JPG up to 8K — clears 300 DPI at standard journal column widths.
8.
Yes. Your style references — journal templates, your own past figures, lab branding — stay confidential.
Never Training Data: Reference uploads are processed for style extraction only. Not fed into model training.
Style Is Extracted, Reference Isn't Kept: We extract numerical style parameters (palette values, stroke weights, font families) and discard the reference image after processing. We don't keep a copy.
Yours to Delete: Remove any uploaded reference from our servers anytime — zero retention.
