PPT Master

PPT Master is an open-source workflow (a "skill") by Hugo He that runs inside agent-capable AI tools and generates natively editable PowerPoint files from source documents or a bare topic. The project describes its output as real PowerPoint shapes, connectors, data-backed charts and tables rather than flattened images. It is MIT licensed, and the latest release listed is v6.7.0.

What It Is

PPT Master is a presentation-generation pipeline driven by an AI agent such as Claude Code, Cursor, Codex CLI or similar tools. Users supply a PDF, DOCX, web page, images or pasted text, and the agent analyses the content, designs the deck, builds SVG pages and exports a .pptx file. The README states that the pipeline runs locally apart from communication with the AI model, and that no specific model or platform is required.

How the Workflow Runs

By default the agent first confirms a design spec (template, format, page count) before authoring; a quick mode skips confirmation and goes straight to generation and export. Output lands in an exports folder as native DrawingML, with page previews written alongside. A flag produces data-backed PowerPoint chart and table objects with Edit Data, saved as a separate file, and formulas compile to editable OMML equations.

Other Routes and Media

Beyond new decks, the README lists distilling reusable brand, style and layout templates from references, filling an existing .pptx with new content while preserving its design, and adding native transitions, animations and narration generated from speaker notes. Images can come from AI generation backends or web image search (Openverse, Wikimedia Commons, and optionally Pexels and Pixabay), with credit lines added when licences require them.

Examples and Setup

The examples gallery lists dozens of generated decks across languages and formats, including Chinese, German, Hebrew, Arabic and Japanese decks, plus square, vertical and A4 canvases such as WeChat and Xiaohongshu formats. Setup requires Python 3.10+ and pip install -r requirements.txt, via git clone, ZIP download or a skill marketplace install. The README advises that a capable large-context model is recommended and that some polishing of the result remains with the user.