Scan Generator

A Flutter application that automatically processes and translates Korean webtoon/manga content by extracting images from web pages, performing OCR text recognition, and translating the text using AI services.

🎯 Project Goal

The Scan Generator is designed to automate the process of translating Korean webtoons and manga by:

1. Web Scraping: Extracting image URLs from Korean webtoon platforms (Naver, Toomics)
2. Image Processing: Downloading and processing comic page images
3. Text Recognition: Using Google ML Kit to detect and extract Korean text from speech bubbles
4. AI Translation: Translating Korean text to English using ChatGPT and DeepL APIs
5. Image Reconstruction: Replacing original Korean text with English translations while maintaining visual quality
6. Batch Processing: Automatically handling multiple pages and chapters

✨ Key Features

🔍 Smart Image Processing

• Automatic Image Detection: Finds and downloads images from supported webtoon platforms
• Split Bubble Detection: Intelligently detects and handles split speech bubbles across page boundaries
• Image Optimization: Converts and optimizes images for processing

📝 Advanced Text Recognition

• Korean OCR: Uses Google ML Kit for accurate Korean text recognition
• Speech Bubble Detection: Automatically identifies and processes text within speech bubbles
• Text Merging: Intelligently combines split text blocks for better translation context

🤖 AI-Powered Translation

• Dual Translation Services:
  • ChatGPT: Context-aware translations with conversation understanding
  • DeepL: High-quality direct translations as fallback
• Context Preservation: Maintains story context across multiple pages
• Smart Caching: Caches translations to avoid redundant API calls

🎨 Visual Text Replacement

• Bubble Recreation: Automatically creates new speech bubbles for translated text
• Font Management: Supports multiple custom fonts for different text styles
• Layout Preservation: Maintains original comic layout and visual flow

🔄 Automated Workflow

• Ticket-Based Processing: Queue-based system for handling multiple translation jobs
• Batch Operations: Processes multiple pages efficiently
• Progress Tracking: Real-time progress monitoring and logging

🏗️ Architecture

Core Services

• ImageService: Handles image processing, OCR, and translation
• ScanService: Manages translation tickets and batch processing
• HttpService: Handles API communications and file downloads
• CacheService: Manages translation caching and file storage

Supported Platforms

• Naver Webtoon: image-comic.pstatic.net
• Toomics: toon.toomics.com

Translation APIs

• OpenAI ChatGPT: Primary translation service with context awareness
• DeepL API: Secondary translation service for direct translations

🚀 Getting Started

Prerequisites

• Flutter SDK (>=2.18.4)
• Android Studio / Xcode for mobile development
• API keys for ChatGPT and DeepL services

Installation

1. Clone the repository
2. Install dependencies:

   flutter pub get

3. Configure API keys in the service files
4. Run the application:

   flutter run

Configuration

Update the following with your API keys:

• ChatGPT API Key: In ImageService._translateWithChatGpt()
• DeepL API Key: In ImageService._translateWithDeepL()

📱 Supported Platforms

• Android: Full functionality with native performance
• iOS: Full functionality with native performance
• Web: Limited functionality (webview restrictions)
• Desktop: Windows, macOS, Linux support

🔧 Technical Stack

Flutter Packages

• google_ml_kit: Text recognition and OCR
• flutter_inappwebview: Web content rendering
• image: Image processing and manipulation
• http: API communications
• flutter_bloc: State management
• path_provider: File system access
• flutter_cache_manager: Caching system

External Services

• Google ML Kit: Korean text recognition
• OpenAI API: Context-aware translation
• DeepL API: High-quality translation
• Custom Backend: Ticket management and batch processing

📊 Workflow

1. Ticket Retrieval: App requests a new translation job from backend
2. Web Scraping: Loads the target webtoon page in webview
3. Image Extraction: Finds and downloads all comic images
4. Image Processing: Splits pages if needed and optimizes for OCR
5. Text Recognition: Extracts Korean text from speech bubbles
6. Translation: Translates text using AI services with context
7. Image Reconstruction: Replaces Korean text with English translations
8. Upload: Sends processed images back to backend
9. Repeat: Continues with next ticket

🎨 Customization

Fonts

The app supports multiple custom fonts located in the assets/ directory:

• Augie
• Another Danger
• BD Cartoon
• CC Astro City
• Zud Juice

Translation Rules

Customize translation behavior by modifying the ChatGPT prompt in ImageService._translateWithChatGpt().

⚠️ Important Notes

• API Rate Limits: Be mindful of API usage limits for translation services
• Copyright: Ensure you have rights to translate and process the content
• Performance: Large batches may take significant time to process
• Network: Requires stable internet connection for API calls

🔒 Security

• API keys are currently hardcoded (consider using environment variables for production)
• All processing is done locally except for translation API calls
• No user data is stored permanently

📈 Future Enhancements

• Support for additional webtoon platforms
• More language pairs for translation
• Improved text bubble detection algorithms
• Batch processing optimization
• User interface improvements
• Translation quality feedback system

🤝 Contributing

This appears to be a personal project for automated webtoon translation. If you're interested in contributing or have suggestions, please feel free to reach out.

📄 License

This project is for personal use. Please respect copyright laws and terms of service of the platforms you're processing content from.