MangaFix

A desktop tool for upscaling entire manga using AI models (Real-ESRGAN, Real-CUGAN) and direct resizing.

Features

• 🚀 AI Upscaling - RealESRGAN x4plus and RealCUGAN x2 support with GPU acceleration
• 📐 Smart Resizing - Automatic resize to target resolution (1080p, 1440p, 4K, or custom)
• 📦 Batch Processing - Process entire manga folders with chapter structure
• 🌙 Modern Dark UI - Clean, dark-themed Material Design interface
• ⚡ GPU Optimized - Configurable tile sizes for different VRAM levels (2GB to 24GB+)
• 💾 System Tray - Minimize to tray for background processing
• 📊 Real-time Progress - Chapter-by-chapter progress tracking with detailed logs

Visual Comparison

🎨 Interactive Before/After Demo →

Drag the slider to compare Original vs RealCUGAN x2 vs RealESRGAN x4plus quality.

Requirements

• Windows 10/11 (64-bit)
• .NET 8.0 Runtime (if not already installed)
• NVIDIA GPU recommended for AI upscaling (optional for resize-only mode)

Installation

1. Download the latest release from Releases
2. Extract the ZIP file
3. Run MangaFix.exe

No installation required - all dependencies are included!

Usage

1. Select Input Folder - Choose a folder with structure: MangaName\Chapter\images.jpg
2. Select Output Folder - Where the processed ZIP files will be saved
3. Choose Target Resolution - 1080p (1920px), 1440p (2560px), 4K (3840px), or Custom
4. Select Upscaling Method:
   • Resize only - No AI, fastest, direct resize (recommended for speed)
   • AI RealCUGAN x2 - Good quality, fast AI upscale (recommended for balanced quality/speed)
   • AI Real-ESRGAN x4 - Best quality, slower AI upscale (recommended for maximum quality)
5. Configure Tile Size - Based on your GPU VRAM (see AI Settings below)
6. Start Processing - Click START and monitor progress

Input Structure

Mangas/
└── MangaName/
    ├── Chapter 1/
    │   ├── 001.jpg
    │   ├── 002.jpg
    │   └── ...
    └── Chapter 2/
        └── ...

Output

Output/
└── MangaName.zip
    ├── 1/  (Chapter 1 images, processed and resized)
    ├── 2/  (Chapter 2 images, processed and resized)
    └── ...

AI Settings

Tile Size (VRAM-based)

• 2 GB VRAM → tile 64
• 4 GB VRAM → tile 128
• 8 GB VRAM → tile 256 (recommended)
• 16 GB VRAM → tile 400
• 24 GB+ VRAM → no tiling

Upscaling Methods

• Resize only - No AI upscale, fastest (direct resize to target resolution)
• AI RealCUGAN x2 - Good quality, fast AI upscale optimized for anime/manga with denoising options
• AI Real-ESRGAN x4 - Best quality, slower upscale using Real-ESRGAN x4plus model

Processing Pipeline

1. Phase 1: Upscaling - AI upscaling (skipped if using Resize only mode)
2. Phase 2: Resizing - Resize to target resolution (automatically upscales or downscales as needed)
3. Phase 3: Zipping - Package chapters into final ZIP file

Building from Source

Prerequisites

• Visual Studio 2022 or .NET 8.0 SDK
• Windows 10/11

Build Steps

cd src
dotnet restore
dotnet build -c Release

Publish Single-File EXE

cd src
dotnet publish -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true -o ../dist

The compiled executable will be in the dist/ folder along with required realesrgan and realcugan binaries.

Credits

MangaFix uses the following open-source projects:

• Real-ESRGAN (BSD-3-Clause) - AI upscaling models
• Real-CUGAN-ncnn-vulkan (MIT) - Anime/manga optimized AI upscaling
• MaterialDesignInXAML (MIT) - UI framework
• Lanczos resampling for high-quality CPU-based resizing

License

MIT License - see LICENSE file for details

Support

If you find this tool helpful, consider buying me a coffee ☕