PaperBanana is an end-to-end AI pipeline for scientific diagrams, flowcharts, architectures, concept maps, schematics, and data plots — with a built-in critic that keeps iterating until the figure is actually good.
flowchart TD R[Receptor] --> T[Transducer] T --> K[Kinase cascade] K --> TF[Transcription] TF --> P[Protein output] K -.-> F[Feedback loop]
Diagrams, flowcharts, architectures, concept maps, schematics, scientific visualizations. One engine, many shapes.
Journal-clean, poster-bold, thesis-formal, training-friendly, SOP-workflow. Or build your own palette + typography.
When numeric fidelity matters, PaperBanana emits exact Chart.js specs — not pixelated renders.
Every figure is scored on faithfulness, simplicity, readability, and aesthetics. Low scores trigger auto-refinement.
Export to HTML, Mermaid source, or PNG. Optimized for paper, poster, and presentation mediums.
Tell the Critic what's wrong. It rewrites the Visualizer prompt and tries again — until the figure is acceptable.
Extracts entities, relationships, and keywords from your brief, image, or dataset.
Designs the layout — top-down, radial, grid — decides sections, items, and annotations.
Applies your chosen palette and typography to the plan. Journal-ready by default.
Emits concrete artifacts — Mermaid for diagrams, Chart.js for plots, SVG for schematics.
Scores the figure on 4 axes. If it falls short, the pipeline refines and tries again.
Minimal black & white for peer-reviewed journals.
High-contrast palette for conference posters.
Neutral greys, serif typography — dissertation-ready.
Warm, approachable palette for e-learning decks.
Procedural colors for SOPs and workflow diagrams.
Soft pastels for concept maps and ideation.