PaperBanana
PaperBanana is an AI-powered academic illustration generator built for researchers who need publication-ready figures without hand-drawing every detail. It converts text descriptions of papers into methodology diagrams, statistical charts, system architectures, flow charts, and poster assets through a multi-agent collaboration pipeline. Instead of offering simple image generation, the tool coordinates planning and visualization agents to produce logically accurate scientific graphics.
At the core of PaperBanana is a closed-loop five-agent architecture. A Planner agent turns textual descriptions into structured visual layouts, while a Visualizer agent renders them with the built-in Nano-Banana-Pro model using precise shapes, connectors, and scientific icons. For statistical plots, PaperBanana generates executable Python Matplotlib code from raw data rather than rendering pixels, which keeps bars, axes, and scales mathematically precise and avoids numerical hallucination. Researchers can also upload rough sketches or whiteboard notes and let the framework apply auto-summarized aesthetic guidelines to refine color, typography, spacing, and overall quality.
PaperBanana is trusted by researchers from institutions such as Stanford, UC Berkeley, CMU, Tsinghua, and SNU. The service offers several credit-based plans with annual image allowances, support for 1K, 2K, and 4K resolutions, and an editing workflow that uses GPT Image 2.5 to change labels, colors, and details while preserving the original figure. It is designed for scientists, students, and authors who want polished figures for top-venue papers, tutorials, lecture slides, and supplementary materials.
