SciFig

SciFig is a generative image tool that turns plain-language descriptions into editable scientific illustrations. SciFig generates labeled scientific illustrations from plain-language descriptions, with model and style choice, conversational editing and PNG or JPG export. It also covers data charts, flowcharts and specialist diagram tools for researchers.

SciFig is a web-based figure generator built for researchers who need scientific illustrations without opening drawing software. The user describes a figure in plain language, naming the components, their arrangement and the exact labels, and the system produces a first draft image. Typical subjects include signaling pathways, laboratory workflows, ecosystems, battery cross-sections and plant anatomy, all shown as examples on the site. In practice the process starts with a prompt box where the user picks a model, aspect ratio and style suited to a slide, poster or manuscript panel. An AI Enhance option can expand a short note into a fuller prompt. The draft then opens in an Illustration workspace, where changes are requested conversationally, a selected area can be redrawn with Region Redraw (a paid feature), and text inside the image can be corrected. Results can be upscaled to 2K, or 4K on Starter and higher plans, then exported as PNG or JPG, or opened in Vector Canvas for vectorized editing. The site is candid that export formats vary by result, so the final file should be checked rather than assumed to be a fully editable vector. The free tier uses GPT Image 2, while paid plans offer six models, including Nano Banana Pro and GPT Image 2.5, with credit costs that differ per model. Beyond pictorial illustration, the platform also includes DataChart for plotting uploaded tables and FlowChart for editable SVG diagrams, plus dedicated tools for graphical abstracts, experimental workflows, circuit diagrams and free body diagrams. SciFig fits scientists, students and educators producing mechanism panels, posters, lecture slides and grant figures. Compared with general image generators, it is tuned to labeled scientific content and keeps editing close to the generation step. Compared with manual tools such as vector editors or icon libraries, it trades precise control for speed, so accuracy checking against source material remains the user's job.