paper2slides

Turn a scientific paper PDF into a presentation slide deck.

An Antigravity / Claude Code skill.

---

Overview

paper2slides is an agent skill that reads a scientific paper, extracts its figures, and produces a ready-to-present Reveal.js slide deck with speaker notes.

It works in two layers:

Layer	Role	Implementation
Extraction	Pull figures from the PDF, segment into panels	Python library (paper2slides)
Curation	Read the paper, select key figures, write slide content and speaker notes	The agent

The extraction step is mechanical — it gives you all the raw material. The agent then does what a human presenter would: reads the paper, decides what matters, and builds a coherent narrative.

Installation

1. Add the skill to your workspace

git clone https://github.com/inhyeoklee/paper2slides-skill.git

# Copy into your workspace
cp -r paper2slides-skill/skill/ YOUR_WORKSPACE/.agent/skills/paper2slides/
cp paper2slides-skill/workflow/paper2slides.md YOUR_WORKSPACE/.agent/workflows/

2. Install the Python library

cd paper2slides-skill/library
pip install -e .

Requires Python 3.9+ and installs PyMuPDF, Pillow, NumPy, Jinja2, and Click.

How It Works

graph LR
    A[Paper PDF] --> B[paper2slides CLI]
    B --> C[Panel Images]
    A --> D[Agent reads text]
    C --> E[Agent curates slides]
    D --> E
    E --> F[Slide Deck + Speaker Notes]

Loading

Workflow

1. Extract — paper2slides paper.pdf -o output/ pulls figures at 4× DPI and segments them into individual panels
2. Read — The agent extracts and reads the paper text to understand the structure, results, and narrative
3. Curate — The agent selects the most important panels, writes slide titles and captions, and composes speaker notes
4. Verify — The agent opens the deck in a browser to check rendering

Usage

Tell the agent:

/paper2slides

Create a presentation from: path/to/paper.pdf

Or use the CLI directly for just the extraction step:

paper2slides paper.pdf -o ./output --serve

What You Get

• 16:9 Reveal.js deck with light/dark themes
• Speaker notes embedded in each slide (press S to view)
• Interactive figures — zoom, pan, and lightbox on any image
• Standalone speaker notes as a separate markdown file
• Settings panel for font, figure scale, and transitions

Project Structure

paper2slides-skill/
├── skill/                       # → .agent/skills/paper2slides/
│   ├── SKILL.md                 #   Agent instructions
│   └── resources/
│       ├── template_shell.html  #   Reveal.js template
│       ├── slides.css           #   Presentation styles
│       └── speaker_notes_template.md
├── workflow/                    # → .agent/workflows/
│   └── paper2slides.md         #   /paper2slides command
├── library/                     # → pip install -e .
│   ├── pyproject.toml
│   └── src/paper2slides/
│       ├── cli.py               #   CLI entry point
│       ├── extractor.py         #   PDF figure extraction
│       ├── segmenter.py         #   Panel segmentation
│       ├── generator.py         #   HTML generation
│       └── templates/
├── LICENSE
└── README.md

Contributing

PRs welcome. Some directions:

• PPTX / Google Slides export
• OCR-based panel label detection
• Multi-paper decks

License

MIT