pdf-to-pptx

A Claude Code skill that converts a PDF into an editable PowerPoint (.pptx) where every piece of text is a real, editable text box — never a flattened image of the page.

The page's graphics (charts, photos, logos) are kept as a background image with the original text erased, so the editable boxes on top don't double up with baked-in text. The hard part — and what most converters get wrong — is the erase step: a naïve converter paints a flat grey rectangle behind every word, which is invisible on a flat page but shows up as an ugly grey box over a photo or gradient. This skill's erase is texture-aware and avoids that.

What it handles

• Multi-column layouts (auto-scaled column splitting)
• Photo / gradient / image-panel backgrounds (feathered, texture-aware erase)
• Font mapping to safe Office fonts and per-cell text colour
• Slide size matched to the original PDF page

Install

Drop the folder into your Claude Code skills directory:

git clone https://github.com/<you>/pdf-to-pptx.git ~/.claude/skills/pdf-to-pptx

Claude will pick it up automatically and invoke it when you ask to turn a PDF into PowerPoint/slides.

Dependencies

The converter relies on:

• poppler — provides pdftoppm for 150-DPI rasterisation
• pdfplumber — text + bounding-box extraction
• Pillow — background image manipulation (the feathered erase)
• python-pptx — .pptx output

# system
sudo apt install poppler-utils      # Debian/Ubuntu; or `brew install poppler` on macOS
# python
pip install pdfplumber Pillow python-pptx

Note: this uses pdfplumber, not PyMuPDF/fitz, for all text extraction.

Usage

Standalone:

python scripts/pdf2pptx.py <input.pdf>

It writes <input_basename>.pptx next to where it's run and prints a report.

Via Claude Code, just ask: "turn this PDF into editable PowerPoint slides."

How it works

See SKILL.md for the full pipeline and the guarantees baked into the script, and references/qa.md for the mandatory QA pass.

License

MIT — see LICENSE.