Banana-slides is a generative AI workflow engine designed to automate the creation and refinement of professional slide decks. By leveraging large language models, the platform transforms raw text, structured outlines, and existing documents into visual presentations. It functions as an automated tool that orchestrates the entire lifecycle of a presentation, from initial content generation and layout design to final export.
The system distinguishes itself through a modular provider abstraction that allows users to integrate various artificial intelligence services for content and image synthesis. It supports iterative workflows where users can refine slide structures, visual styles, and descriptions using natural language commands. The platform maintains state-based project persistence, enabling users to manage version history, perform batch operations, and regenerate specific slides without discarding the entire project.
Beyond core generation, the project provides a comprehensive suite of tools for managing visual assets, including the ability to apply custom branding, generate data visualizations, and synthesize narrations. It handles complex document parsing to refurbish existing files and supports multiple export formats, including editable slides and PDFs. The system is built to handle large-scale tasks through asynchronous processing and provides hooks for external integrations.