PPTAgent

Report Abuse

Basic Information

PPTAgent is a research and engineering project that automatically generates and evaluates presentation slides from input documents. The repository implements a two-phase pipeline inspired by human presentation creation: an Analysis Phase that extracts patterns and learns from reference presentations, and a Generation Phase that produces structured outlines and visually cohesive slides. It also introduces PPTEval, an evaluation framework that assesses generated presentations across multiple quality dimensions. The README references a paper, a dataset, demonstration materials and documentation, and provides a pre-built Docker image for quick start. The project includes case study examples and sample slide images to illustrate outputs. The codebase is intended to support automated slide generation research and practical presentation synthesis workflows rather than being a simple text-to-slides converter.

Links

Categorization

App Details

Features
The project highlights dynamic content generation that integrates text and images into slides and a smart reference learning mechanism that leverages existing presentations without manual annotation. It follows a two-phase workflow with separate analysis and generation steps to improve structure and visual coherence. PPTEval offers a comprehensive evaluation across content accuracy and relevance, design aesthetics and consistency, and coherence of information flow. The repository provides a demo video, case study images, documentation, a cited research paper, and a linked dataset. A pre-built Docker image is provided for easier setup. The README includes visual workflow diagrams and examples demonstrating how the system analyzes references and produces slide outputs.
Use Cases
PPTAgent automates the time-consuming task of creating presentations by converting documents into structured slide outlines and visually coherent slides, which can speed up content production and improve consistency. Its Analysis Phase helps the system learn layout and stylistic patterns from real presentations, reducing manual tuning and enabling outputs that resemble human-created slides. PPTEval gives objective, multi-dimensional feedback on content relevance, visual design, and narrative coherence, which supports iterative improvement and benchmarking. Provided demo materials, documentation, dataset and a Docker image lower the barrier to experimentation and reproduction, making the repository useful for researchers, developers, and teams exploring automated presentation generation and quality assessment.

Please fill the required fields*