GitHub - genmini-ai/OpenCanvas: Generate stunning presentations from any topic or PDF in minutes - with AI evaluation and self-evolution

AI-powered presentation generation from topics or PDFs. Beautiful slides in minutes, not hours.

✨ Key Features

• 🎨 Dual Output Modes - HTML slides or PNG images with Gemini AI
• 📄 Flexible Input - Generate from topics or PDF documents
• 🎯 Smart Figure Matching - Auto-extracts and matches figures to slides (Image mode + PDF)
• 📊 Quality Evaluation - AI-powered assessment and improvement
• 🚀 REST API - Full programmatic access

🚀 Quick Start

Installation

git clone https://github.com/genmini-ai/OpenCanvas.git
cd OpenCanvas
pip install -e .

Setup

cp .env.example .env
# Add your ANTHROPIC_API_KEY (required for both modes)
# Add your GEMINI_API_KEY (required for Image mode)

Get API keys:

• Anthropic: console.anthropic.com
• Gemini: aistudio.google.com

Generate Your First Presentation

🔥 Image Mode (PNG slides with AI-generated visuals)

# From a topic
opencanvas generate "quantum computing" \
  --output-format image \
  --theme minimalist

# From a PDF (with automatic figure extraction & matching)
opencanvas generate "https://arxiv.org/pdf/2412.06769" \
  --output-format image \
  --theme minimalist

Available themes: minimalist (tested), academic, corporate, creative, tech, cartoon Example presentations: Minimalist theme | Cartoon theme PDF

Image mode: AI-generated slide with automatically extracted and matched figures from research paper

HTML Mode (Default)

# From a topic
opencanvas generate "AI in healthcare" --purpose "conference talk"

# From a PDF
opencanvas generate "https://arxiv.org/pdf/2412.06769" --purpose "conference talk"

HTML mode: Clean, professional slides with embedded styles

📖 Output Modes

🔥 Image Mode

• Output: Individual PNG slides + compiled PDF
• Best for: High-quality visuals, research presentations with figures
• Features: AI-generated designs, automatic figure extraction from PDFs, intelligent figure-to-slide matching
• Time: ~3 minutes for 10-12 slides
• Requires: GEMINI_API_KEY

Learn more about Image Mode →

HTML Mode

• Output: Single HTML file with embedded styles
• Best for: Quick presentations, web sharing, easy editing
• Features: Clean layouts, multiple themes, fast generation
• Convert to PDF: opencanvas convert slides.html

📁 Output Structure

HTML Mode:

output/topic_20241213_225914/
├── slides/
│   └── slides.html          # Single HTML file
└── sources/
    └── source_content.txt

Image Mode:

output/topic_20241213_225914/
├── slides/
│   ├── slide_001.png
│   ├── slide_002.png
│   └── ...
├── extracted_images/        # (PDF input only)
│   ├── figure_1.png
│   └── figure_captions.json
├── sources/
│   ├── blog_content.txt
│   └── slide_blueprints.json
└── presentation.pdf         # Auto-compiled

📚 Documentation

• Installation Guide - Detailed setup
• CLI Reference - All commands
• Image Mode Guide - Figure matching & PNG slides
• API Guide - REST API
• Evolution System - Auto-improvement

🎯 Common Use Cases

# HTML: Quick presentation from topic
opencanvas generate "machine learning basics" --purpose "lecture"

# HTML: Research paper to slides
opencanvas generate "paper.pdf" --purpose "seminar"

# Image: High-quality presentation with figures
opencanvas generate "paper.pdf" \
  --output-format image \
  --purpose "conference talk"

# Complete pipeline with evaluation
opencanvas pipeline "research topic" --purpose "conference" --evaluate

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

# Development setup
git clone https://github.com/genmini-ai/OpenCanvas.git
cd OpenCanvas
pip install -r requirements-all.txt
python run_tests.py

📄 License

MIT License - see LICENSE file for details.

🙏 Built With

• Anthropic Claude - AI generation
• Google Gemini - Image generation
• Docling - Figure extraction
• FastAPI - REST API

⭐ Star this repo if you find it useful!

📚 Full Documentation | 🐛 Report Issues | 💬 Discussions