PPT Master is an open-source generator that turns a document, or just a topic, into a native PowerPoint file rather than a web slideshow or a PDF. It is aimed at people whose slides have to survive being opened, edited and reshuffled by someone else: analysts, researchers, lecturers, consultants, and anyone who sends decks into an organisation that runs on Office. The project is written in Python, MIT licensed, and has collected roughly 59,000 stars and 4,600 forks on GitHub since it appeared in December 2025.
What it does
You point it at source material and it writes a .pptx. By the project's own account, the output includes:
- native shapes instead of flat text boxes, plus slide transitions and animations
- charts and tables built from real data, generated on demand rather than pasted in as pictures
- audio narration produced from the speaker notes
- support for your own .pptx templates, so existing corporate styling carries over
The framing both the author and the video use is that the model reasons through the whole argument first and designs slides afterwards, so the structure follows the content instead of a theme you fill in. The video shows a research paper becoming native diagrams and native formulas. That is what the demo shows, not an independent measurement, and there is no published benchmark attached to it.
How it works
The tool runs locally and writes the finished file on your machine. It does not ship with a model of its own: you connect one, which is either a paid cloud API key or a model running on your own hardware. That split matters for cost. The software is free and open source under MIT, and nothing about it needs a subscription, but the intelligence does: either you pay a provider per token, or you pay in local compute. The README names a set of sponsors, most of them API services (Kimi, PackyCode, APIKEY.FAN, RunAPI, APIMart), which is how the author keeps the project itself free.
Because the generation targets PowerPoint's own object model, the result is meant to be editable afterwards in the same way a hand-built deck is: you can move a shape, retype a label, or adjust a chart's numbers. The repository is bilingual (English and Chinese READMEs) and is mirrored on AtomGit alongside GitHub.
Getting started
The repository is github.com/hugohe3/ppt-master. Two things are needed before the tool is useful: a Python environment, and access to an AI model, which is the part that costs money unless you run something locally. Tagged releases exist, and the README's version badge points at the GitHub Releases page, so you can pin a version rather than track the default branch.
On hardware, be aware that the project does not list requirements: no operating system minimum, no memory or VRAM figure, no chip requirement. If you go the local-model route, the real constraint is whatever model you choose, not PPT Master itself. You will also want PowerPoint, or an app that reads .pptx faithfully, to see transitions and animations as intended. For the exact install commands, read the README on the repository page; this entry does not quote a command line, because the setup section was not part of the material it was written from.
When to use it / when not
The main catch is honesty about maturity: the project's stated direction is that each release gets its output closer to a deck a person built natively in PowerPoint, which is a goal rather than a finished claim. Expect to review and fix slides, especially dense ones.
Use it instead of asking Claude or ChatGPT for a slide outline when you need the actual file, with native shapes, a slide master and charts tied to data, and not a wall of bullet text you then rebuild by hand. Use it instead of a hosted deck generator when the material cannot leave your machine, since the document processing happens locally and only the model call goes out (and not even that, with a local model).
Don't use it if you want a click-and-go web service with no key to configure and no Python to install, if your slides live in Google Slides or Canva and never need to be a PowerPoint file, or if you are unwilling to spend either API credits or local GPU time on every deck.
PPT Master is worth taking seriously if the .pptx format is a requirement in your work rather than an inconvenience, because that is the narrow thing it does that most AI slide tools do not: it writes PowerPoint's own objects instead of pictures of slides. The star count says plenty of people recognised that gap. Treat the setup as a developer-grade task and the output as a strong first draft, and it saves the dull half of deck building; expect a polished, hands-off product and it will disappoint.