HTML slide authoring + PDF export pipeline, driven by AI agents.

Write a slide in slides/, run one command, get a PDF.

Works with any AI agent that can read CLAUDE.md — Claude Code, Cursor, Windsurf, GitHub Copilot, and others.

1. Write your slides as a single HTML file in slides/
2. Run ./run.sh to convert to PDF via headless Chromium
3. Pick up the result from output/

slides/my-talk.html  →  output/my-talk.pdf

# install deps (first time only)
python3 -m venv venv
source venv/bin/activate
pip install playwright
playwright install chromium

# convert a single file
./run.sh slides/my-talk.html

# convert everything in slides/
./run.sh

make-ppt/
├── slides/          # put your HTML slides here
├── assets/          # background & decoration images
├── fonts/           # local font files (Tossface, Freesentation, Wanted Sans)
├── output/          # generated PDFs (git-ignored)
├── convert.py       # Playwright-based converter (don't touch)
├── run.sh           # entry point
├── CLAUDE.md        # authoring rules for AI agents
└── DESIGN.md        # design system reference

Each .html file in slides/ is a standalone deck. One <div class="slide"> = one PDF page.

<div class="slide slide-title">
  <!-- title page -->
</div>

<div class="slide slide-content">
  <!-- content page -->
</div>

Fixed canvas: 1920 × 1080px (16:9).

Background images and right-side decorations live in assets/.

See slides/asset-test.html for a visual reference of every asset.

All fonts are bundled locally so PDF export works offline.

Pick one body font per deck — don't mix.

MIT