🎮 METranslator

عربي

METranslator is a powerful, privacy-focused, and local-first AI translation tool specifically designed for games and general text. It leverages state-of-the-art neural machine translation models to provide high-quality translations directly on your machine.

---

✨ Key Features

• 🏠 100% Local Processing: Your data never leaves your computer. All translations are performed locally using powerful AI models.
• 🚀 Hardware Acceleration: Supports both CPU and NVIDIA GPU (CUDA) for lightning-fast translation speeds.
• 🧠 Multiple AI Backends:
  • OPUS-MT: High-performance models for specific language pairs.
  • MADLAD-400: A massive multilingual model for high-quality variety.
  • mBART-50: Versatile many-to-many translation.
• 🛠️ Built-in Model Converter: Easily convert models to CTranslate2 or ONNX formats to optimize performance for your specific hardware.
• 📦 Portable Environment: Automated setup that creates a self-contained Python environment without messing up your system settings.
• 🌍 Multilingual UI: Native support for Arabic, English, French, German, Spanish, Italian, Portuguese, Russian, Chinese, Japanese, Korean. (Note: Interface translations were generated by AI. If you find any errors, please report them to me).

🔌 Integration with LunaTranslator

METranslator is fully compatible with LunaTranslator, a popular visual novel translator.

To use METranslator as a translation source in LunaTranslator:

1. Download the adapted configuration file: LunaTranslator Selfbuild Script
2. Place this file in your LunaTranslator userconfig folder.
3. Select Custom Translation in the LunaTranslator settings.

(This integration relies on a modified selfbuild.py originally from the LunaTranslator project).

---

� Screenshots

Main Interface

Settings Configuration

Download & Converter Management

Virtual Environment Setup

Usage Instructions

---

�🛠️ Prerequisites

• OS: Windows 10/11 (64-bit recommended).
• GPU (Optional): NVIDIA GPU with CUDA support for significantly faster translation.
• RAM: 8GB minimum (16GB+ recommended). Note: ONNX conversion requires at least 16GB of RAM.
• Disk Space: Approximately 500MB for the core app, plus 1GB-4GB depending on the AI models you choose to download.

---

🚀 Quick Start

1. Download:

   • Download the latest version from Releases.
   • Extract the archive to your desired location.

   OR

2. Clone the Repository:

   git clone https://github.com/Marwan087/METranslator.git
   cd METranslator

3. Run the Installer/App: Launch gui_installer.py (or the compiled METranslator.exe). The installer will:

   • Setup a local Python runtime.
   • Install all necessary dependencies (PySide6, requests, etc.).

4. Launch the App: Once installed, run METranslator.exe .

5. Setup in App:

   • Step 1: Environment Setup: Choose your device type (CPU or GPU). Click Start to install the specific AI libraries (Note: GPU setup may download ~3GB of CUDA libraries).
   • Step 2: Model Acquisition: In Download & Convert, choose your language pair or a multilingual model (MADLAD/mBART).
   • Step 3: Optimization (Optional): Convert models to CTranslate2 or ONNX for maximum performance on your hardware.
   • Step 4: Launch: Go to Settings to select your model and device, then click Run Server ▶ to start translating!

---

�️ How to Build

If you want to build the executable (.exe) files yourself, follow these steps:

1. Requirements

• Python Version: Python 3.13.5 (Ensure it's added to your PATH).
• Compression Tool: 7-Zip must be installed (for creating the final archive).
• Build Dependencies: Install the required libraries using:

  pip install -r requirementsbuild.txt

  (Libraries include: pyinstaller, cx_freeze, certifi, and requests).

2. Build Process

Run the provided batch script to automate the build process:

build_release.bat

This script will:

1. Compile the main application using cx_Freeze.
2. Package the uninstaller using PyInstaller.
3. Organize all necessary files into a temporary folder.
4. Create a compressed METranslator.7z archive (The final ready-to-use package).
5. Clean up temporary build directories.

---

🏗️ Technology Stack

• UI Framework: PySide6 (Qt for Python)
• Backend API: FastAPI & Uvicorn
• AI Engine: Hugging Face Transformers, CTranslate2, and ONNX Runtime
• Models: Credits to Helsinki-NLP, Google (MADLAD), and Facebook (mBART).

---

�📜 License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for more details.

---

🤝 Contributing

Contributions are welcome! If you have suggestions for improvements or encounter bugs, please open an issue or submit a pull request.

---

✉️ Contact

Author: Marwan Alhamaty
Email: marwanalhamaty87@gmail.com
Project Link: https://github.com/Marwan087/METranslator

---

Developed with ❤️ for the gaming and translation community.