The PPTX graveyard is full of tools that promised "AI-generated slides" and delivered glorified screenshots — this one outputs actual editable PowerPoint objects.
The Summary
- PPT Master is an open-source workflow that runs inside AI IDEs (Claude Code, Cursor, VS Code + Copilot) and generates natively editable PowerPoint files from PDFs, DOCX, URLs, or Markdown — real shapes and text boxes, not images.
- The demo shows a 12-page deck generated end-to-end from a single WeChat article URL using Claude Opus 4.7, with every element clickable and editable in PowerPoint.
- Unlike competing tools that export slides as images or locked formats, this outputs actual OOXML — the native PowerPoint file format — with support for animations and transitions.
The Signal
Most "AI presentation tools" are PowerPoint karaoke. They generate slides that look like slides but act like JPEGs. You can't click into a chart to change the data. You can't adjust a text box without regenerating the whole deck. PPT Master solves the format problem that's plagued every AI slide tool since they started appearing: it outputs actual PowerPoint objects, not pictures of PowerPoint objects.
The architecture is unusual. PPT Master isn't a standalone app or web service. It's a workflow — Hugo He calls it a "skill" — that runs inside AI coding environments like Claude Code or Cursor. You drop in source material (a PDF, a URL, a Markdown file) and chat with the AI: "make a deck from this." The AI follows the PPT Master workflow to generate a .pptx file on your local machine. The workflow is open source, so you can see exactly what it's doing: parsing your content, structuring slides, generating OOXML (Office Open XML, the format PowerPoint actually uses), and writing a file that PowerPoint can natively open and edit.
"If a file can't be opened and edited in PowerPoint, it shouldn't be called a PPT."
This matters because the AI presentation space is crowded with tools that generate slide decks as locked PDFs or image exports. Those tools optimize for speed and polish, but they break the edit loop. If you want to change one number in a chart, you're regenerating the whole thing or opening Photoshop. PPT Master's output is fully editable:
- Real text boxes that you can click and retype
- Real shapes with adjustable colors, borders, and positions
- Real charts built from data tables, not bitmaps
- Animations and transitions exported as OOXML, not embedded video
The cost model is transparent. The tool itself is free and open source. The only expense is the API calls to the underlying LLM (Claude Opus 4.7 in the demo). You pay OpenAI, Anthropic, or whoever directly. No markup. No per-seat license. No "credits." This is consistent with the broader pattern of AI tooling moving toward open workflows that run on commodity infrastructure rather than walled SaaS platforms.
The Implication
If you've been waiting for AI slide generation to hit "actually useful" — meaning you can hand it to a non-technical colleague and they can edit it without calling you — this is the closest I've seen. The catch is setup friction: you need Python installed, you need an AI IDE, and you need to be comfortable chatting with an agent to direct the workflow. That's not "open PPT Master dot com and upload your file" simple. But it's also not coding. It's conversational automation with full output control.
Watch for this pattern to spread. Agents that output natively editable artifacts (not exports, not images, not proprietary formats) are the unlock for real delegation. If the output is locked, you're still doing the work. If the output is editable, the agent just became your junior designer.