Manga PDF Colorizator 🎨

An advanced AI-powered tool for colorizing black and white manga PDFs while preserving existing colored pages and text quality.

🌟 Features

• Smart Colorization: Automatically detects and preserves existing colored pages
• Text Preservation: Enhanced text clarity and sharpness in speech bubbles
• Border Optimization: Automatic white border removal for cleaner output
• AI Upscaling: Optional RealESRGAN upscaling for higher quality
• GPU Acceleration: CUDA support for faster processing

📋 Requirements

System Requirements

• Python 3.7+
• CUDA-capable GPU (optional, for GPU acceleration)
• 8GB RAM minimum (16GB recommended)
• 2GB free disk space

Dependencies

# Core dependencies
PyMuPDF==1.21.1
numpy>=1.21.0
Pillow>=9.0.0
opencv-python>=4.5.0
torch>=1.9.0
colorama>=0.4.4
basicsr>=1.4.2  # For RealESRGAN

🚀 Installation

1. Clone the repository:

git clone https://github.com/ahorasoft/manga-colorizator.git
cd manga-colorizator

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

3. Install dependencies:

pip install -r requirements.txt

4. Download required models:

# Create directories
mkdir -p networks denoising/models

# Download colorization model
wget https://github.com/ahorasoft/manga-colorizator/releases/download/v1.0/generator.zip -O networks/generator.zip

# Download denoising models
wget https://github.com/ahorasoft/manga-colorizator/releases/download/v1.0/denoising_models.zip
unzip denoising_models.zip -d denoising/models/

# Download RealESRGAN model (optional, for upscaling)
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus_anime_6B.pth -P upscaling/

💻 Usage Examples

Command Line Usage

# Basic usage
python inference_pdf.py "manga.pdf"

# Advanced usage with all options
python inference_pdf.py \
    -sp 1 -ep 140 \        # Start page 1, end page 140
    -z 1.5 \               # Zoom factor 1.5x
    -ai \                  # Enable AI upscaling
    -pt 0 \               # Disable text preservation
    -ci 1 \               # Normal color intensity
    -cf 1.2 \             # Increased contrast
    -sf 1.0 \             # Normal sharpness
    "path/to/manga.pdf"

# Real example with enhanced contrast and sharpness
python inference_pdf.py -sp 14 -ep 17 -z 1.5 -ai -pt 0 -ci 1 -cf 1.5 -sf 1.2 "manga/chapter.pdf"

# Process specific chapter with GPU
python inference_pdf.py \
    -sp 45 -ep 65 \       # Process pages 45-65 only
    -g \                  # Enable GPU
    -s 768 \              # Higher resolution
    -ds 25 \              # Denoising strength
    "manga/chapter5.pdf"

📸 Example Results

Left: Original black & white manga page. Right: AI-colorized result with enhanced contrast (cf=1.5) and sharpness (sf=1.2)

Command used for this example:

python inference_pdf.py -sp 14 -ep 17 -z 1.5 -ai -pt 0 -ci 1 -cf 1.5 -sf 1.2 "manga/chapter.pdf"

Key settings:

• AI upscaling enabled (-ai)
• No text preservation (-pt 0) for artistic consistency
• Normal color intensity (-ci 1)
• Enhanced contrast (-cf 1.5)
• Subtle sharpness increase (-sf 1.2)

🛠️ Configuration Options

Option	Default	Description
-o, --output	input_colorized.pdf	Output PDF path
-g, --gpu	False	Enable GPU acceleration
-s, --size	576	Processing size (must be multiple of 32)
-nd, --no_denoise	False	Disable denoising
-ds, --denoiser_sigma	25	Denoising strength (1-50)
-z, --zoom	1.0	Output zoom factor (0.1-4.0)
-ai, --ai_upscale	False	Enable RealESRGAN upscaling
-pt, --preserve_text	0.5	Text preservation strength (0.0-1.0)
-ci, --color_intensity	1.0	Color intensity (0.5-2.0)
-cf, --contrast_factor	1.2	Contrast enhancement (0.5-2.0)
-sf, --sharpness_factor	1.0	Sharpness enhancement (0.0-2.0)

📁 Project Structure

manga-colorizator/
├── networks/
│   └── generator.zip          # Colorization model
├── denoising/
│   └── models/               # Denoising models
├── upscaling/
│   └── RealESRGAN_x4plus_anime_6B.pth  # Upscaling model
├── inference.py              # Single image colorization
├── inference_pdf.py          # PDF processing
└── colorize.py              # CLI interface

🔍 Troubleshooting

Common Issues

1. Model Loading Errors

   Error: Could not find generator.zip
   Solution: Ensure all models are downloaded to correct directories
   

2. CUDA Out of Memory

   # Reduce processing size
   colorizer = PDFColorizer(
       use_gpu=True,
       size=384  # Default is 576
   )

3. Blurry Text

   # Increase text preservation
   colorizer = PDFColorizer(
       preserve_text=0.8,
       denoiser_sigma=15
   )

4. Over-processed Images

   # Reduce enhancement factors
   colorizer = PDFColorizer(
       contrast_factor=1.1,    # Subtle contrast increase
       sharpness_factor=1.0,   # No sharpness change
       color_intensity=1.0     # Normal colors
   )

📧 Contact

• Website: http://www.ahorasoft.com
• Email: support@ahorasoft.com
• Issues: GitHub Issues

---

Made with ❤️ by Ahorasoft