PaperBanana

What is PaperBanana

PaperBanana is an AI-powered academic illustration generator that automates the creation of publication-ready figures from text descriptions. It utilizes a multi-agent framework to produce methodology diagrams, statistical charts, infographics, and other scientific visualizations.

How to use PaperBanana

1. Enter a prompt: Describe the desired academic figure in the text area. Templates are available for common structures like "Transformer Architecture" or "RAG System Pipeline".
2. Select Style: Choose a visual style from the dropdown menu (e.g., Methodology Diagrams, Statistical Plots).
3. Configure Options: Adjust Aspect Ratio and Resolution as needed.
4. Generate: Click the "Generate" button to create the illustration.
5. Preview: View the generated image in the preview pane.

Features of PaperBanana

• Automated Illustration Generation: Creates publication-ready academic illustrations from text.
• Multi-Agent Framework: Employs a closed-loop system with five agents (Retriever, Planner, Stylist, Visualizer, Critic) for accurate and polished results.
• Methodology Diagrams: Generates model architectures, algorithm flows, and system pipelines.
• Accurate Statistical Plots: Produces charts using executable Python Matplotlib code to ensure numerical precision and avoid hallucinations.
• Aesthetic Enhancement: Refines rough sketches and existing diagrams to meet publication standards.
• Educational Infographics: Creates clear and scientifically rigorous infographics for educational purposes.
• Style Transfer: Applies aesthetic guidelines from reference libraries for top-venue compliance.
• Open Source: Available under CC BY-SA 4.0 license with code and datasets on GitHub.

Use Cases of PaperBanana

• Generating methodology diagrams for research papers.
• Creating accurate statistical charts and plots from data.
• Producing educational infographics to explain complex concepts.
• Enhancing and polishing hand-drawn sketches into professional illustrations.
• Developing visual assets for academic posters and presentations.

FAQ

What is PaperBanana?

PaperBanana is an agentic framework designed to automate the generation of publication-ready academic illustrations, including methodology diagrams, statistical plots, and educational infographics. It orchestrates five specialized AI agents in a collaborative pipeline to ensure accuracy, faithfulness, and visual quality.

How does PaperBanana's multi-agent framework work?

PaperBanana coordinates five specialized agents: Retriever (locates references), Planner (translates text to layouts), Stylist (synthesizes aesthetics), Visualizer (renders using Nano-Banana-Pro or code), and Critic (self-reflects and corrects). This closed-loop architecture ensures faithfulness, precision, and reliability.

How does PaperBanana ensure data accuracy in statistical plots?

For statistical charts, PaperBanana generates executable Python Matplotlib code from raw data, ensuring mathematical precision for bars, data points, and scales, thus avoiding numerical hallucinations.

What is Nano-Banana-Pro?

Nano-Banana-Pro is a specialized image generation model used by PaperBanana's Visualizer Agent, excelling at synthesizing complex shapes, connectors, and scientific icons for methodology diagrams.

How is PaperBanana evaluated?

PaperBanana was evaluated using PaperBananaBench on 292 NeurIPS 2025 test cases, showing consistent outperformance against baselines like GPT-Image and Paper2Any in faithfulness, conciseness, readability, and aesthetics.

Is PaperBanana open source?

Yes, PaperBanana is fully open source under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0), with its codebase, model weights, and benchmark publicly available on GitHub.

What inputs does PaperBanana need to generate a diagram?

Typically, it requires a Visual Intent (description), Source Context (paper sections), and a Figure Caption. For statistical plots, Raw Data in JSON or CSV format is also needed.

Can PaperBanana improve my existing hand-drawn sketches?

Yes, PaperBanana's Aesthetic Enhancement feature applies auto-summarized guidelines to refine color schemes, typography, and overall quality of hand-drawn drafts.

What types of academic figures can PaperBanana generate?

It supports methodology diagrams, statistical charts, educational infographics, poster and conference slide assets, and aesthetic refinement of existing sketches.

How does PaperBanana compare to DALL-E or other image generators?

Unlike general image generators that may produce logical or numerical errors, PaperBanana's structured multi-agent pipeline, code-based rendering for charts, and critic agent ensure scientific faithfulness and accuracy.