PPTAgent is an automated system for generating presentations from documents, designed for researchers and developers working with AI-driven content creation. It addresses the challenge of producing high-quality, visually cohesive presentations by mimicking human creation processes and includes a robust evaluation framework, PPTEval, to assess content, design, and coherence.

How It Works

PPTAgent employs a two-phase approach: an Analysis Phase to extract and learn patterns from reference presentations, followed by a Generation Phase that develops structured outlines and creates visually integrated slides with text and images. This method aims to leverage existing presentation styles without manual annotation for improved quality and coherence.

Quick Start & Requirements

• Install: Pre-built Docker image available.
• Requirements: Docker.
• Links: Paper, Documentation, Demo Video

Highlighted Details

• Dynamic content generation with integrated text and images.
• Smart reference learning from existing presentations.
• Comprehensive quality assessment via PPTEval (Content, Design, Coherence).

Maintenance & Community

This project is actively maintained and welcomes contributions via issues, pull requests, and discussions.

Licensing & Compatibility

The README does not specify a license.

Limitations & Caveats

The project is presented as a research artifact, and its readiness for production environments or commercial use is not detailed.