English | 한국어 | Français | 简体中文 | 日本語 | Português Brasileiro
Made with ❤️ by Un Community led by SitriSaleos
UnComicTranslate is a community-driven fork focused on providing open access to powerful comic translation capabilities using State-of-the-Art (SOTA) Large Language Models.
Our Mission:
- Open Access: Keep translation capabilities accessible to everyone using their own API keys, without forced credit systems
- Enhanced Features: Expand functionality with user flexibility and advanced export options
- Community-Driven: Maintain an active development focused on user needs
Automatic comic translation tool supporting multiple languages and LLM providers. Translates to and from English, Korean, Japanese, French, Simplified Chinese, Traditional Chinese, Russian, German, Dutch, Spanish, and Italian. Also supports translation to Turkish, Polish, Portuguese, and Brazilian Portuguese.
Table of Contents:
- Python 3.12
- Git
- uv package manager
-
Install Python 3.12 from python.org (check "Add python.exe to PATH")
-
Install Git from git-scm.com
-
Install uv from docs.astral.sh
-
Clone and setup the repository:
git clone https://github.com/SitriSaleos/UnComicTranslate
cd UnComicTranslate
uv sync- (Optional) For NVIDIA GPU support:
uv pip install onnxruntime-gpugit pull
uv add -r requirements.txt --compile-bytecodeLaunch the GUI:
uv run main.py-
CBR Files: If using CBR files, install WinRAR or 7-Zip and add the installation folder to your system PATH. See instructions for Windows, Linux, or Mac
-
Font Support: Ensure your selected font supports characters of the target language
-
Manual Mode: When automatic mode encounters issues (undetected text, incorrect OCR, insufficient cleaning), use Manual Mode to make corrections. Undo the image and toggle Manual Mode
-
Automatic Processing: Once an image is processed, it loads in the Viewer while other images continue translating in the background
-
Navigation:
- Ctrl + Mouse Wheel to zoom (or vertical scroll)
- Trackpad gestures supported
- Left/Right arrow keys to navigate between images
Configure providers in Settings -> Credentials -> Select Supplier.
UnComicTranslate supports over 15+ Providers (including Online APIs and Local models):
- OCR: Runs locally and is completely free by default using Manga-OCR, Pororo, or PPOCR. Optional cloud providers (Google/Azure) are available if needed.
- Translation (LLM):
- Gemini Flash (2.0/2.5/3): Highly recommended as Google provides approximately 1,500 free requests per day (RPD) for these models.
- Other providers: Supports OpenAI (GPT-4o), Anthropic (Claude), Deepseek, Groq, HuggingFace, Ollama (Local), 9Router, and more.
Uses RT-DETR-v2 model trained on 11k comic images (manga, webtoons, western comics) for speech bubble and text detection, followed by algorithmic segmentation.
Default providers:
Optional (requires API key):
Removes detected text using:
Supports GPT-4o, GPT-4o-mini, DeepL, Claude-3, Gemini-2.5, Yandex, Google Translate, and Microsoft Azure Translator. Full page context provided to LLMs for better translations. Optional image context available.
Renders translated text in bounding boxes from detected speech bubbles.
Export translated projects in web-ready format:
- Clean inpainted images without text
- Structured JSON files with coordinates, original text, and translations
- Perfect for building interactive comic viewers or mobile apps