PPT Master – AI‑generated native PowerPoint decks

What it does

PPT Master is a Python‑based tool that lets a large‑language‑model (LLM) read a source document (PDF, DOCX, web page, etc.) and output a real .pptx file. The generated deck is not a flat image – it contains PowerPoint’s native objects:

• slide masters and layouts
• editable shapes, connectors and text boxes
• data‑backed charts and tables
• native pictures, fills and effects
• optional transitions, animations and narration

Because the output uses PowerPoint’s own object model, you can open the file in PowerPoint and continue editing exactly as if you had built the slides yourself.

How it works (high‑level)

1. Agent‑capable AI tool – you run the workflow inside any chat/IDE that can act as an agent (read/write files, run commands). The README recommends Claude Code, Kimi K3, or any similar model.
2. Prompt & model – you tell the agent “make a deck from this PDF”. The model parses the source, extracts key points, decides a slide structure, and may request image generation (e.g., gpt‑image‑2 or Gemini‑nano‑banana‑2.1).
3. Python pipeline – the model’s instructions are executed by the PPT Master Python scripts, which:
   • convert the source to plain text (using built‑in parsers or optional Pandoc for legacy formats)
   • build a PowerPoint file with the python‑pptx library, populating native shapes, charts, tables, etc.
   • write the final .pptx to disk.
4. Local execution – all processing happens on your machine; only the LLM calls go out to the chosen API provider. No PowerPoint subscription is required.

Main features

Feature	What you get
Native PowerPoint objects	Slides contain real shapes, charts, tables, and slide‑master inheritance – fully editable in PowerPoint.
Multiple input formats	PDF, DOCX, HTML, EPUB, IPYNB and, with optional Pandoc, older formats like .doc, .odt, .tex.
Template extraction & reuse	Derive a brand/template deck from existing slides and reuse it for new content.
Style‑preserving fill	When updating an existing .pptx, the tool keeps the original design while swapping in new text/images.
Optional transitions/animations/narration	Add native PowerPoint transitions, simple animations, and audio narration as a post‑process step.
Agent‑agnostic	Works with Claude, Kimi, GPT, Gemini, etc., as long as the model can read/write files.
Transparent cost	The only expense is the LLM API usage; the software itself is MIT‑licensed and free.
Local data	Your source documents never leave your computer (except for the LLM calls).

Getting started (quick summary)

1. Install Python 3.10+.
2. Clone the repo (or download the ZIP) and run pip install -r requirements.txt.
3. Choose an agent – e.g., Claude Code, Kimi K3, or any CLI/IDE that supports file‑read/write agents.
4. Run the skill – in the agent’s chat window type a command like make a deck from <path‑to‑file.pdf>; the agent will invoke the PPT Master pipeline and produce output.pptx.
5. Open the resulting file in PowerPoint and edit as needed.

The README also provides:

• a Windows‑specific installation guide,
• a skill‑marketplace manifest for one‑click installation in Claude Code,
• a live demo site with several example decks, and
• a roadmap/FAQ outlining future native‑PowerPoint capabilities.

Who it’s for

• Professionals who need fast, editable slide decks from reports, research papers, or brand assets.
• Developers or power‑users comfortable with a Python environment and an AI‑agent chat tool.
• Anyone who wants to keep their data local and avoid vendor‑locked PowerPoint‑generation services.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.