What is reference to figure?

Reference to figure is LabFig's style-transfer mode. You upload two images — your own content figure and a reference whose look you want — and reference to figure repaints your content in the reference's visual language: its palette, typography, arrow style, and box treatment. You get your data in a new look, not a copy of the reference scientific figure. For a worked example, read how authors make AI figures for machine learning papers .

Does it copy the data or just the look?

Just the look. LabFig takes the reference's colors, line weights, typography, and layout conventions and nothing else. The content, panels, numbers, and labels all come from the figure you upload as your own — the reference never contributes its subject matter, only its visual style.

What makes a good reference image?

A clean, high-resolution scientific figure with a clear and consistent visual language works best for reference to figure — distinct colors, readable typography, and obvious arrow and box conventions. Busy screenshots or low-resolution scans give the model less to learn from. One strong reference beats several conflicting ones.

Can I match my own previous figure for a multi-panel paper?

Yes — that is the most common use. Upload your approved Figure 1 as the reference, then upload each new panel as your content. Every panel inherits the same palette, typography, and arrow style, so the whole multi-panel paper or thesis stays visually consistent without manual restyling.

What about the copyright and licensing of a reference I borrow?

Use only references you have the right to use — your own figures, your lab's work, or material you are licensed to reuse. Style transfer borrows visual language rather than copying content, but you remain responsible for respecting the source's license and citing it where required. When in doubt, reference your own past work.

How is reference to figure different from text to figure?

Text to figure builds a scientific figure from a written description with no image input. Reference to figure starts from images — your content plus a style reference — and its job is to match a look you already have. Use text to figure to create from scratch, and reference to figure to make new content fit an existing visual standard. Once the look is matched, you can polish the result with the figure enhancer or fine-tune it on the vector canvas before export.