Skip to Content

PosterGen: Academic Poster Creation with Multi-Agent AI

Transforming Research Papers into Professional Posters Through Intelligent Multi-Agent Collaboration
Zhilin Zhang

Creating compelling conference posters is a challenge for any researcher. You have to accurately and compelling decide which content and how it will be presented. After months of rigorous research, writing, and peer review, scientists face the daunting task of distilling their complex findings into a visually appealing poster that can capture attention in a crowded conference hall. This process, traditionally requiring design expertise that most researchers lack, often results in cramped, aesthetically poor presentations that fail to do justice to groundbreaking research.

Y-Research-SBU

Y-Research-SBU

Organization

PosterGen

Official Code for PosterGen
141
8
1
8 Network
3 Subscribers
Python
agentlarge-language-modelvision-language-models

PosterGen is an AI First approach to academic poster generation that harnesses the power of multi-agent large language models (LLMs) to transform research papers into professionally designed posters. Developed by a collaborative team from Stony Brook University, New York University, University of British Columbia, Zhejiang University, and UCLA, this innovative system represents departure from how we typically approached scientific communication design, all thanks to advancements in AI.

The Problem: Design Meets Science

Creating an effective academic poster requires a unique blend of skills that few researchers possess. The poster must accurately represent complex research findings while adhering to design principles that are appealing and readable. Traditional approaches to automated poster generation have fallen short because they focus primarily on content extraction without considering the aesthetic and design principles that make posters truly effective to a broad audience.

The challenge extends beyond simple layout choices: posters must balance information density with visual clarity, maintain consistent typography and color schemes, and guide the viewer's eye through a logical narrative flow. Most existing solutions produce posters that, while technically correct, require extensive manual refinement to meet professional standards.

The Solution: Multi-Agent Intelligence

PosterGen tackles this through an innovative multi-agent framework that attempts to mirror the workflow of professional poster designers. Rather than treating poster creation as a single-step process, the system breaks down the task into specialized components, each handled by a dedicated AI agent with specific expertise.

The framework consists of six specialized agents working in orchestrated collaboration. The Parser Agent extracts and structures content from research papers using the advanced Marker PDF processing library. The Curator Agent then organizes this content into a narrative-based storyboard, following the "And, But, Therefore" (ABT) structure that helps create compelling scientific narratives.

The Layout Agent, working in tandem with a Balancer Sub-Agent, calculates precise positioning and spacing while optimizing column utilization to prevent overflow. This is followed by the Color Agent that generates cohesive color schemes based on institutional branding from affiliation logos, and the Font Agent that applies professional typography and keyword highlighting. Finally, the Renderer composes the final poster using the python-pptx library.

Why I Like It

What sets PosterGen apart is its deep understanding of design principles embedded within the AI agents themselves. Unlike systems that simply arrange content spatially, PosterGen incorporates fundamental design concepts like alignment, proximity, repetition, and contrast directly into its decision-making process. The system creates aesthetically coherent, professionally designed presentations that require minimal human intervention.

The project's approach to evaluation is particularly compelling. Rather than relying solely on traditional metrics, the team developed a Vision-Language Model (VLM)-based evaluation rubric that measures layout balance, readability, and aesthetic coherence. This innovative evaluation method demonstrates the project's commitment to understanding and quantifying design quality, not just content accuracy.

Key Features: Professional Design Made Accessible

PosterGen's standout features reflect its deep integration of design principles with advanced AI capabilities. The system automatically extracts figures, tables, and text from research papers while maintaining their semantic relationships and visual integrity. Its three-column layout engine ensures natural reading flow and optimal space utilization, crucial for conference poster effectiveness.

The intelligent color scheme generation is particularly innovative, analyzing institutional logos to extract brand-appropriate color palettes that ensure visual consistency with the researcher's affiliation. The typography system applies hierarchical font choices and strategic keyword highlighting to create clear information hierarchies that guide reader attention.

Perhaps most importantly, PosterGen produces both high-resolution PNG images suitable for printing and fully editable PowerPoint files, giving researchers flexibility for final adjustments. The system supports custom dimensions with intelligent aspect ratio optimization, accommodating various conference requirements from ISO A-series to custom poster sizes.

Under the Hood: Engineering Excellence

Built on the LangGraph framework, the system implements a robust multi-agent workflow with precise state management and error handling.

The codebase is organized into logical modules within the src directory, with separate packages for agents, configuration, layout processing, state management, and workflow orchestration. 

The choice to use Python 3.11 ensures compatibility with the latest language features while maintaining stability for the extensive dependency stack that includes PyTorch, transformers, and computer vision libraries.

# Example of the workflow pipeline structure
graph = StateGraph(PosterState)

# Add specialized agents
graph.add_node("parser", parser_node)
graph.add_node("curator", curator_node) 
graph.add_node("color_agent", color_agent_node)
graph.add_node("layout_optimizer", layout_optimizer_node)
graph.add_node("font_agent", font_agent_node)
graph.add_node("renderer", renderer_node)

# Define workflow sequence
graph.add_edge(START, "parser")
graph.add_edge("parser", "curator")
graph.add_edge("curator", "color_agent")
# ... additional edges defining the pipeline

The system's configuration architecture, centered around poster_config.yaml, allows for fine-tuning of layout parameters, typography settings, and color generation algorithms without modifying core code. This design philosophy makes the system adaptable to different institutional requirements and design preferences.

Integration with multiple LLM providers (OpenAI, Anthropic, Google) through environment-based configuration ensures flexibility in model selection and provides fallback options for reliability. The web interface, built with React and TypeScript, offers an intuitive drag-and-drop experience that makes the powerful backend accessible to researchers without technical expertise.

Use Cases: Transforming Academic Communication

PosterGen addresses a wide range of academic communication scenarios beyond traditional conference presentations. Research institutions can use the system to create consistent, branded poster templates for graduate student symposiums and departmental showcases. The system's ability to extract institutional colors and maintain design consistency makes it valuable for creating cohesive visual identities across research groups without stressing out the presenters.

In educational settings, PosterGen can help students learn effective scientific communication by providing professionally designed examples that demonstrate proper information hierarchy and visual design principles. The editable PowerPoint output allows educators to modify generated posters for teaching purposes, highlighting specific design decisions and their impact on communication effectiveness.

For researchers in developing regions or smaller institutions with limited design resources, PosterGen democratizes access to professional-quality poster design. The system levels the playing field, ensuring that groundbreaking research isn't overshadowed by poor presentation quality at international conferences.

Community: Building the Future of Scientific Communication

The PosterGen project has quickly established an active community around scientific communication and design automation. The GitHub repository serves as a hub for collaboration, with clear contribution guidelines and an engaged development team that actively responds to issues and feature requests.

The project maintains multiple communication channels, including a WeChat group for Chinese-speaking users and a Discord community for broader international collaboration. This multi-platform approach reflects the project's commitment to global accessibility and inclusive community building.

A live demo on Hugging Face Spaces allows researchers to experiment with the system without installation, lowering barriers to adoption and enabling rapid feedback from the academic community. The project's comprehensive documentation and example datasets in the data directory facilitate easy onboarding for new users.

Usage & License Terms

PosterGen is released under the MIT License, providing maximum flexibility for both academic and commercial use. This permissive licensing allows researchers to freely use, modify, and distribute the software without restrictive copyleft requirements. Users can incorporate PosterGen into their research workflows, modify the code for institutional-specific needs, and even develop commercial applications based on the codebase.

The MIT License specifically grants rights to use, copy, modify, merge, publish, distribute, sublicense, and sell copies of the software, provided that copyright notices and license terms are preserved. This licensing choice reflects the academic community's commitment to open science and collaborative development.

Impact Potential: Democratizing Design Excellence

PosterGen embodies a fundamental shift toward democratizing design in academic communication. By embedding professional design principles into AI agents, the system makes high-quality visual communication accessible to researchers regardless of their design background or institutional resources.

The project's impact extends beyond individual poster creation to broader questions about AI-assisted creativity and the automation of design processes. The multi-agent approach demonstrates how complex creative tasks can be decomposed into specialized components, each leveraging AI capabilities while maintaining human-like design sensibilities.

As the system evolves, its influence on academic conference presentation standards could be significant. By raising the baseline quality of poster presentations, PosterGen may drive overall improvements in scientific communication, making research more accessible and engaging for diverse audiences.

The open-source project ensures that its benefits extend beyond the original development team, creating opportunities for international collaboration and adaptation to different cultural and institutional contexts. This global accessibility aligns with the broader mission of open science and equitable access to research tools.

Conclusion: The Future of Scientific Communication

PosterGen stands as a testament to the power of thoughtful AI application in solving real-world academic challenges. By combining deep technical expertise with an understanding of design principles and user needs, the project creates genuine value for the research community. The system doesn't just automate poster creation; it elevates the entire process, ensuring that brilliant research receives the visual presentation it deserves.

For researchers tired of struggling with design software or accepting subpar poster quality, PosterGen offers a glimpse into a future where AI handles the technical complexity of design while preserving the human creativity and insight that drive scientific discovery. This is open-source innovation at its finest: solving practical problems while advancing the state of the art in AI-assisted creativity.

Whether you're a graduate student preparing for your first conference or a seasoned researcher looking to streamline your presentation workflow, PosterGen deserves a place in your toolkit. The project represents not just a technological achievement, but a step toward a more accessible and visually compelling future for scientific communication.


Authors:
Zhilin Zhang
PosterGen: Academic Poster Creation with Multi-Agent AI
Joshua Berkowitz September 22, 2025
Views 1474
Share this post