nabz0r / mac-local-translator

Local translation app for Mac using speech recognition and offline translation

Real-time, fully offline, privacy-friendly conversational translation 🔒

🌟 Key Features

• Fully local: No internet connection required 📶
• Private & Secure: Your conversations never leave your device 🔐
• Bidirectional: Natural translation between two speakers 👥
• Powerful: Models optimized for Intel and Apple Silicon Macs ⚡
• Intuitive: Elegant, conversation-centric interface 💬

🔍 Overview

Mac Local Translator allows you to translate conversations in real-time without relying on cloud services. The app uses advanced local models for speech recognition and translation, ensuring your privacy while delivering excellent performance.

flowchart LR
A[User speech A] -->|Microphone| B[Speech recognition]
B -->|Transcribed text| C[Translation]
C -->|Translated text| D[Text-to-speech]
D -->|Audio| E[User B hears]
E -->|Answers| F[Reverse cycle]
F -->|Microphone| B

Loading

🚀 Quick Install

Requirements

• macOS 12.0 or newer
• 8 GB RAM minimum (16 GB recommended)
• 5 GB of available disk space

Method 1: Install from DMG

1. Download the latest version from the releases page
2. Mount the DMG file and drag the application into your Applications folder
3. Launch the application

Method 2: Build from source

# Clone the repository
git clone https://github.com/nabz0r/mac-local-translator.git
cd mac-local-translator

# Install dependencies and download models
make download-models

# Build the application
make build

# Launch the application
make run

🏛️ Architecture

The application is built on a modular architecture for easy maintenance and extension.

releasesarchitecture_overviewpage

🎯 User Guide

1. Language Selection: Choose the source and target languages ​​in the bar tools
2. Start: Press the record button or use Cmd+Space
3. Speak: Speak clearly into your microphone (the audio level indicator will help you)
4. Listen: The translation will be displayed and pronounced automatically
5. Reply: Your interlocutor can respond in their language and the cycle continues

🛠️ Technologies used

• User interface: SwiftUI
• Speech recognition: Whisper.cpp (optimized version of OpenAI Whisper)
• Translation: Optimized LibreTranslate/Argos Translate models
• Text-to-speech: AVSpeechSynthesizer with quality voices

🗣️ Supported languages

• 🇫🇷 French
• 🇬🇧 English
• 🇪🇸 Spanish
• 🇩🇪 German
• 🇮🇹 Italian
• 🇵🇹 Portuguese

More languages ​​are under development. Check out the roadmap for more information.

🤝 Contribution

We welcome:

• 💻 Network Engineers
• 👀 Research Scientists
• 🌎 Cloud Architects
• 🤖 AI/ML Specialists

📱 Contact

• 📧 Email: [e-mail]
• 🐙 GitHub: @nabz0r

📄 License

MIT License - Innovation without Boundaries

🙏 Thanks

• OpenAI Whisper for the speech recognition model
• LibreTranslate for the translation models