Transform academic papers into adorable manga with AI

Prism is an open-source tool that converts complex academic papers (PDFs) into engaging, easy-to-understand manga-style comics. Powered by Gemini's image generation capabilities, it makes learning fun and accessible for everyone.

• PDF to Manga Conversion - Upload any academic paper and get a beautifully illustrated manga
• Multiple Art Styles - Choose from 3 unique manga themes:
  • Kumomo - Original cute characters (kumo, nezu, papi) with consistent design across all panels
  • Chiikawa - Cute, simple characters with soft pastel colors (Nagano style)
  • Studio Ghibli - Dreamy watercolor atmosphere (Spirited Away style)
• Character Consistency - Reference images ensure characters look the same throughout the entire manga
• AI-Powered Storyboarding - Intelligent breakdown of complex concepts into visual panels with proper story ordering
• Multi-Language Support - Generate manga in English, Chinese (中文), or Japanese (日本語)
• CJK Text Optimization - Dynamic batch sizing for clear, readable Chinese/Japanese text
• Real-time Generation Progress - Watch your manga come to life panel by panel

┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Upload PDF │ ──▶ │   Analyze   │ ──▶ │ Storyboard  │ ──▶ │  Generate   │
│             │     │  (English)  │     │  + Translate│     │ Manga Panels│
└─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘

1. Upload - Drop your PDF academic paper
2. Analyze - AI extracts and understands the content in English (for accuracy)
3. Storyboard - Creates a visual narrative, then translates dialogue to target language
4. Generate - Gemini renders each manga panel with consistent character designs

For optimal quality, Prism uses a three-step process:

1. English Analysis - Technical content is analyzed in English for accuracy
2. English Storyboard - Panels and dialogue are created in English first
3. Translation - Final dialogue is translated to the target language (zh-CN, ja-JP)

This ensures technical terms are understood correctly before translation.

• Python 3.11+
• Node.js 18+
• OpenRouter API key (for Gemini access)

# Clone the repository
git clone https://github.com/raucvr/Prism.git
cd Prism

# Backend setup
cd backend
pip install -r requirements.txt

# Frontend setup
cd ../frontend
npm install

1. Copy the example config file:

cp config/api_config.yaml.example config/api_config.yaml

2. Set your OpenRouter API key in config/api_config.yaml:

api_key: "your-openrouter-api-key"

Or use environment variable:

# Windows
set OPENROUTER_API_KEY=your-openrouter-api-key

# Linux/macOS
export OPENROUTER_API_KEY=your-openrouter-api-key

# Terminal 1 - Start backend
cd backend
python -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload

# Terminal 2 - Start frontend
cd frontend
npm run dev

Open http://localhost:3000 in your browser.

prism/
├── backend/
│   ├── engines/           # AI engine implementations
│   │   ├── base.py        # Abstract base classes
│   │   └── openrouter.py  # OpenRouter API wrapper
│   ├── services/
│   │   ├── pdf_parser.py      # PDF text extraction
│   │   ├── storyboarder.py    # Storyboard generation
│   │   └── manga_generator.py # Image generation with character consistency
│   ├── routes/            # API endpoints
│   └── main.py            # FastAPI application
├── frontend/
│   ├── src/
│   │   ├── app/           # Next.js app router
│   │   ├── components/    # React components
│   │   └── store/         # Zustand state management
│   └── package.json
├── config/
│   ├── api_config.yaml.example  # Config template
│   └── character_images/        # Reference images for Kumomo theme
│       ├── kumo.jpeg
│       ├── nezu.jpeg
│       └── papi.jpeg
└── README.md

For the Kumomo theme, Prism ensures character consistency by:

1. Reference Images - Loading character design images for each API call
2. Explicit Mapping - Telling the model exactly which image corresponds to which character
3. Low Temperature - Using temperature=0.3 to reduce randomness
4. Negative Prompts - Explicitly forbidding character design deviations

For CJK languages (Chinese, Japanese), Prism dynamically adjusts batch size:

• Long dialogue (>200 chars): 1 panel per batch
• Medium dialogue (>100 chars): 2 panels per batch
• Short dialogue: 4 panels per batch (2x2 grid)

This ensures text remains readable even with complex characters.

All panels are sorted by panel_number after generation to ensure the story flows correctly, even if the AI generates them out of order.

curl -X POST http://localhost:8000/api/manga/from-pdf \
  -F "file=@paper.pdf" \
  -F "theme=kumomo" \
  -F "language=zh-CN"

• FastAPI - High-performance Python web framework
• httpx - Async HTTP client
• PyMuPDF - PDF parsing
• Pillow - Image processing

• Next.js 14 - React framework with App Router
• TailwindCSS - Utility-first CSS
• Zustand - State management
• Framer Motion - Animations
• react-dropzone - File upload

We welcome contributions!

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

• Batch processing for multiple papers
• Custom character upload
• Panel layout customization
• Export to PDF/EPUB
• More art styles
• Mobile app

This project is licensed under the MIT License - see the LICENSE file for details.

• OpenRouter - API gateway for AI models
• Chiikawa - Inspiration for cute art style

Made with ❤ by the Prism Team

Report Bug · Request Feature