Skip to content

NeTRuNNeRGLiTCH/manga-cleaner

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🌌 MANGA CLEANER v3.0.0: The Professional Studio Update

High-Fidelity AI-Powered Scanlation Restoration Suite

Download v3.0.0 Python Version

Manga Cleaner v3.0.0 represents a complete paradigm shift in scanlation technology. Moving from the legacy "script-based" architecture of v2.0.1 to a modular "Studio" engine, this update delivers unprecedented speed, precision, and hardware efficiency.


🖥️ Studio Preview (v3.0.0 UI)

Manga Cleaner v3.0 UI Featuring the new Obsidian Dark Theme and Hardware Telemetry.


🚀 The v2.0.1 ➔ v3.0.0 Evolution

The jump to v3.0.0 isn't just an update—it's a total re-engineering of the restoration pipeline.

💎 Resolution-Invariant Tiling Engine (NEW)

In v2, tiling often resulted in slight quality degradation or blurring at the seams.

  • v3 Improvement: Our new Dynamic ROI Engine maintains 1:1 pixel integrity. By utilizing "Snap-to-8" reflection padding, the engine processes ultra-high 4K+ resolutions without downscaling, ensuring the output is as sharp as the original scan. No more Gaussian blurring needed—the seams are mathematically perfect.

📉 Massive GPU Footprint Optimization

We have successfully decoupled the runtime from unnecessary bloat.

  • Size Reduction: The GPU package has been slashed from 2.6GB down to 900MB (download size) (and from 370 to 270 for CPU version) through aggressive dependency pruning and specialized ONNX runtime stripping, making it easier to download and share.

🧠 Advanced AI Core

  • Detection: Upgraded to a high-frequency heatmap OCR for pinpoint accuracy on Japanese, Korean, and Chinese text bubbles.
  • Inpainting: Fully migrated to a specialized ONNX-quantized LaMa core for 40% faster execution on both CPU and GPU.

📸 Before & After Comparison

Original Scan AI Restored (v3.0.0)
Before 1 After 1
Before 2 After 2

📦 Version Comparison: GPU vs. CPU

Feature ⚡ CUDA GPU Edition ☁️ Standard CPU Edition
Ideal For High-end NVIDIA Gaming PCs Laptops & Integrated Graphics
Package Size ~900MB (Optimized) ~260MB (Ultra-Lite)
Cleaning Speed Instant / Real-time 3-8 seconds per page
Hardware NVIDIA GPU (CUDA Required) Any modern x64 Processor

🎨 Professional Studio Features

  • Batch Engine: Process entire chapters in one click. Load -> Auto-Scan -> AI Clean -> Export.
  • Selection Toolkit: Added professional Rectangular Selection and Lasso Tools for manual mask refinement.
  • Photoshop® Bridge: Direct COM Interop. Cleaned pages are injected directly into Adobe Photoshop as layered documents.
  • Integrated Help System: A built-in manual and shortcut legend for a zero-friction learning curve.

⌨️ Key Shortcuts

  • [D] Auto-Scan Text | [C] Launch LaMa Clean | [Space] Pan Image
  • [B] Brush | [R] Rectangle | [L] Lasso | [Shift] Toggle Erase And Print Mode
  • [Ctrl+Z] Undo Image Change (LaMa) | [Alt+Z] Undo Manual Mask

🛠️ Technical Stack (For Developers)

  • Language: Python 3.12.7 (Strict OOP Architecture)
  • GUI: PySide6 (Qt6) with Obsidian Dark styling.
  • Processing: OpenCV 4.11 + NumPy 1.26.
  • Async Logic: Multi-threaded QThread workers to maintain 60FPS UI responsiveness during AI tasks.
  • Telemetery: Real-time RAM and VRAM monitoring via psutil and onnxruntime provider telemetry.

🗺️ Roadmap: The Future of Manga Cleaner

  • v4.0.0 (Planned): AMD GPU Support via DirectML Execution Providers and some of the community suggestions

📜 Community & Usage Policy

Copyright © 2024-2026 NeTRuNNeRGLiTCH. All Rights Reserved.

This project is built to empower the scanlation community. I want this tool to help you work faster and more professionally.

  • Profit from your Work: You are 100% free to use this tool to clean manga for your scanlation groups, even if those groups accept donations or work on paid platforms. The output you create is yours.
  • Open for Learning: You are welcome to study the source code, fork the repo, and modify it for your own personal use or to suggest improvements to this project.
  • Strict Commercial Restriction: You are strictly prohibited from selling this software, its source code, or any modified versions of it. You cannot "re-skin" this app and put it behind a paywall.
  • Attribution: While not required, a shout-out to this repo helps the project grow!

⚙️ Development Installation (Source Only)

  1. Clone & Environment:
    git clone https://github.com/NeTRuNNeRGLiTCH/manga-cleaner.git
    cd manga-cleaner
    python -m venv venv
    .\venv\Scripts\activate

About

tool for automating the cleaning process in manga

Topics

Resources

Stars

42 stars

Watchers

3 watching

Forks

Contributors

Languages