# slide-maker: turn a paper, a repo, a doc, or just a topic into a native PPTX you can present

<p align="center"> <a href="README_CN.md"><strong>简体中文</strong></a> </p>

<p align="center"> <a href="LICENSE"><img alt="License: MIT" src="https://img.shields.io/badge/License-MIT-yellow.svg"></a> <img alt="Codex" src="https://img.shields.io/badge/Codex-supported-111827"> <img alt="Claude Code" src="https://img.shields.io/badge/Claude_Code-supported-5b5bd6"> <img alt="Output: editable PPTX" src="https://img.shields.io/badge/output-native_editable_PPTX-0f766e"> </p>

<p align="center">_{Free and open source, built by <a href="https://addsum.top/"><strong>Addsum</strong></a>}</p>

> **The slide-maker that reads your actual work, never invents a number, ships fully-editable native PowerPoint, and won't hand it over until an independent critic signs off.**

Chat with it in Codex or Claude Code. It isn't one prompt guessing at slides: a small team of specialized agents reads your paper / repo / doc (or researches the topic when you have none), plans the story, designs each slide around it, builds a real `.pptx`, and puts it through an independent review before you ever see it.

Most AI-PPT tools race to look pretty in seconds. slide-maker optimizes for the four things that actually matter when the deck is *yours to defend*:

- 🔍 **Reads your source — doesn't invent it.** Every number and figure traces back to your material; it won't make up a statistic to fill a slide. (The failure mode of every "expand a topic" tool — one popular assistant printed *43% growth* where the real figure was *12%*.) - ✏️ **A real PowerPoint, not a screenshot.** Every text box, shape, native chart and equation is a click-to-edit object — *nothing* flattened to an image. (Many "export to PPTX" tools quietly turn a third of your slides into uneditable pictures.) - 🧑‍⚖️ **Reviewed before you see it.** A non-negotiable actor–critic loop: a *separate* critic tries to break the deck —