Paper Banana is an AI workflow for researchers that converts paper text, method sections, and figure captions into structured figure drafts with reference-guided planning and iterative refinement.
What is Paper Banana?
Paper Banana is an online AI tool specifically designed for academics and scientists who need to generate research figures, scientific diagrams, and method visuals from written descriptions. It takes inputs such as paper summaries, methods paragraphs, figure captions, or workflow notes and produces structured figure drafts in PNG format at resolutions up to 4K. The platform runs entirely in the browser at PaperBanana.art and is built by the team behind the Paper Banana project, with public technical materials on GitHub and arXiv supporting the workflow.
Key Features
- Text-to-figure drafting – Paste a method section, figure caption, or workflow description and receive a structured diagram draft. The system analyzes the text to plan the figure layout.
- Multiple candidates and iterative refinement – Generate several candidate figure directions at once, then refine the best option through a critic loop. This reduces rework when comparing layout approaches.
- High-resolution export – Export figures at 1K, 2K, or 4K resolution (credit cost: 5, 10, or 20 credits respectively). Output format is PNG, with aspect ratios 4:3 and 16:9 supported.
- Templates for common diagrams – Choose from ready‑made templates such as “Transformer Architecture” or “GAN Training Pipeline” to jump‑start method figures and system diagrams.
- Reference image and paper upload – Upload a reference image or a full paper PDF (Basic and Pro plans) to automatically retrieve the visual style and apply it to generated figures.
- Intelligent aesthetic refinement – Advanced plans include a refinement step that improves figure composition, labeling, and visual clarity before final export.
- Credit‑based usage – Each image generation consumes credits depending on resolution (1K = 5 credits, 2K = 10, 4K = 20). Credits roll over in annual subscriptions.
Who is it for?
- Academic researchers – Turn dense method text or figure captions into clear, publication‑ready diagrams for papers, posters, and slide decks.
- Graduate students – Quickly generate schematic overviews of their thesis work or lab protocols without learning complex illustration software.
- Lab managers and principal investigators – Create standardized method figures for grant proposals, lab websites, or team communication using consistent visual styles.
What can you do with it?
- Generate method figures from text – Paste a pipeline description from a paper draft and receive a structured flow diagram with labeled modules. Iterate on layout and style until it matches the intended narrative.
- Compare multiple diagram directions – Generate 3–5 candidate layouts for the same method, then pick the most effective one for your audience.
- Export high‑resolution figures for publication – Set resolution to 2K or 4K, apply style refinement (on eligible plans), and download a PNG ready for inclusion in a journal submission or conference poster.
How does it work?
The workflow follows three steps: (1) Paste your method text, figure caption, or workflow notes into the prompt field. Optionally select a mode (e.g., “Methodology Diagrams”) and a template. (2) Generate multiple candidate figure drafts with a single click. (3) Refine the best candidate using iterative prompts and reference images, then export the final figure at your chosen resolution. A history tab (View history) lets you revisit previous generations.
