AI That Outputs Real PowerPoint, Not a Stack of PNGs

What it does

PPT Master is a Python workflow that runs inside AI IDEs like Claude Code or Cursor. You feed it source material and it produces a genuine .pptx built from native DrawingML shapes, editable text boxes, charts, slide transitions, and entrance animations. It can also generate speaker notes and turn them into audio narration, all while running locally on your machine.

The interesting bit

The author, a finance professional who regularly edits decks, built this because he was tired of AI tools that output image-based slides you can’t tweak. The project treats itself as a “harness, not a complete agent”—it owns the workflow, but the quality ceiling is set by whichever model you plug in, with Claude’s large context window currently delivering the best results.

Key highlights

• Outputs real, element-by-element editable PowerPoint files, not image-packed PPTX containers
• Runs fully offline apart from AI model calls; no third-party server touching your documents
• Supports custom .pptx templates, native animations, and AI-generated audio narration from speaker notes
• Works across multiple AI IDEs and models (Claude, GPT, Gemini, Kimi)
• Transparent cost: the open-source tool is free; you only pay for the AI model tokens you consume

Caveats

• The README is upfront that this is a tool, not a “wishing well”—cheap models leave more cleanup work, and top-tier output requires top-tier models like Claude with ~1M token context
• There is a learning curve if you’re new to Python or AI agents; the author explicitly states the project is partly an excuse to teach those skills

Verdict

Worth exploring if you regularly build presentations and want AI to handle the tedious first draft while keeping full editability. Skip it if you need a one-click, zero-effort slide generator or refuse to touch an AI IDE.