Webtoon Translator

A Python-based tool for automatically translating Korean webtoons to English. This tool uses OCR to detect text in speech bubbles, translates the text using OpenAI's GPT-3.5-turbo (the cheapest model), and renders it back onto the image while preserving the original artwork.

Features

• Automatic text detection in speech bubbles using EasyOCR
• Korean to English translation using OpenAI GPT-3.5-turbo (much better quality than MarianMT)
• Intelligent sound effect translation (e.g., "어허회" → "sob", "퍽" → "whack")
• Context-aware translations that maintain plot coherence
• Text removal and inpainting to preserve artwork
• Support for both single panels and long webtoon strips
• Automatic panel splitting for long webtoon images
• Translation caching to reduce API costs

Installation

1. Clone the repository:

git clone https://github.com/cphanhth/Webtoon-Translator.git
cd Webtoon-Translator

2. Create a virtual environment and install dependencies:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

3. Set up OpenAI API key (required for translation):

python setup_openai.py

Or manually create a .env file with:

OPENAI_API_KEY=your_openai_api_key_here
OPENAI_MODEL=gpt-3.5-turbo
OPENAI_MAX_TOKENS=1000
OPENAI_TEMPERATURE=0.3

Usage

Command Line Interface

Single Panel Translation

python src/main.py input_panel.png output_panel.png

Long Webtoon Translation

python src/main.py long_webtoon.png translated_webtoon.png --mode webtoon

With Custom Font

python src/main.py input.png output.png --font path/to/font.ttf

With OpenAI API Key (if not in .env)

python src/main.py input.png output.png --openai-key YOUR_API_KEY

Python API

from src.main import WebtoonTranslator

# Initialize translator
translator = WebtoonTranslator()

# Translate single panel
translator.translate_panel("input_panel.png", "output_panel.png")

# Translate long webtoon
translator.translate_webtoon("long_webtoon.png", "translated_webtoon.png")

Testing

Test the OpenAI integration:

python test_openai.py

Cost Estimation

Using GPT-3.5-turbo (the cheapest model):

• Input tokens: $0.0015 per 1K tokens
• Output tokens: $0.002 per 1K tokens
• Typical cost per panel: ~$0.0001 - $0.0005
• 1000 panels: ~$0.10 - $0.50

The system includes caching to avoid re-translating identical text.

Configuration

You can customize the OpenAI settings in your .env file:

• OPENAI_MODEL: Model to use (default: gpt-3.5-turbo)
• OPENAI_MAX_TOKENS: Maximum response length (default: 1000)
• OPENAI_TEMPERATURE: Response creativity 0.0-1.0 (default: 0.3)

Requirements

• Python 3.7+
• OpenCV
• NumPy
• Pillow
• EasyOCR
• OpenAI (for translation)
• python-dotenv (for configuration)

License

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