PPT Master

hugohe3/ppt-masterReviewed by AI

What it does

PPT Master is a workflow (a "skill") you install into an AI assistant such as Claude Code, Cursor or Codex. You put a PDF, a Word file or just a topic in a folder, ask for a deck in the chat, and it writes a .pptx with real PowerPoint shapes, charts, tables, transitions and speaker notes, so you can keep editing it in PowerPoint. It can also reuse your own template or fill an existing deck. Everything runs on your computer; you install Python 3.10 or newer, the assistant, and the project files. The AI model is what costs money.

How easy it is to install

C · install 1 or 2 tools · Windows

Windows C
macOS C
Linux C

C on Windows, Mac and Linux: none of the 38 releases is an installer, so you install Python 3.10 or newer, install an AI assistant, download the project (git clone or a ZIP) and run `pip install -r requirements.txt`; after that you only type in a chat window. Windows takes extra steps for PATH and the execution policy, which the project covers in its own docs/windows-installation.md.

Numbers on GitHub

Stars: 71,657on Oct 31, 2026
Downloads: 165,6698,218 this week
Growth: +2,747stars in 5 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Python, Other tools
To build from source: Python
Disk space: 100 MB–300 MB · estimate
Accounts or keys: An AI model through your assistant (Claude, Kimi, GPT, Gemini), OpenAI, Google Gemini or another image model, for AI-generated pictures, Pexels or Pixabay, free keys for better photo search
Cost: Free, API costs extra · MIT and free, with "no PPT subscription on top" as the README puts it; what you pay is the AI model that drives it, plus any image generation you turn on. The README recommends a large-context model (Kimi K3 or Claude) and gpt-image-2.5-sunburst or Gemini for pictures, and says cheaper models leave more work for you.
License: MIT (Permissive)
Can replace: Gamma, 365 Copilot
Last update: Oct 30, 2026

Risk signals (6)

Attentionreadme steers you to api resellers it earns from
The README's sponsor section and .env.example "Sponsor relay shortcuts" push paid API resellers with referral links and promo codes, for example "https://www.packyapi.ai/register?aff=ppt-master" ("enter the promo code ppt-master during recharge to get 10% off"), "https://apikey.fan/register?aff=PPT-MASTER" and "https://runapi.host/register?aff=WMLJ". These relays resell access to Claude, GPT and Gemini "as low as 7% of official rates", which is not the providers' own channel.
AI review
Infopdf support brings in an agpl library
README Acknowledgments: "The optional PDF converter depends on PyMuPDF, which is licensed under AGPL-3.0 rather than MIT; it is only needed for PDF sources. See conversion.md before redistributing a bundle that includes it."
AI review
Infofound pictures can require a credit line
README step 5: web search considers "CC0, Public Domain, Pexels / Pixabay no-attribution licenses, CC BY, and CC BY-SA" and "Executor adds a small inline credit whenever the selected image requires attribution", with `--strict-no-attribution` as the way out. Slides may therefore carry credits you must keep.
AI review
Infothe author warns results depend on the model
README, "This is a tool, not a wishing well": "don't expect a finished, perfect deck in one shot" and "The cheaper the model, the more there is to do"; the recommended setup is a roughly 1-million-token context model plus paid image generation.
AI review
Infohuge star count with almost no open issues
The repository was created on 2025-12-10 and shows 71,322 stars and 5,562 forks with a single open issue, so public discussion of problems is thin for a project of this size.
AI review
Infoupdate runs a script that pulls code
README "Updating Later": for clone installs you run `python3 skills/ppt-master/scripts/update_repo.py`, which "pulls the latest version and syncs Python dependencies when requirements.txt changes". So what you run can change without you reviewing it.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether decks come out like the published examples when you use a cheaper model or your own template.
How many tokens, and how much money, one deck costs on your plan.
Whether the Windows guide is enough on a machine with no development tools.
What the scripts send to the image, search and speech services you enable.
Whether the "fill an existing .pptx" route really keeps untouched pages byte-for-byte as the README claims.