Real-Time Voice Translation Application for macOS

A powerful desktop application that captures live audio from video calls or system audio, converts speech to text using OpenAI Whisper, and translates it in real-time to your desired language.

🎯 Features

• Live Audio Capture: Capture from microphone or system audio (video calls, music, etc.)
• Offline Speech Recognition: Uses OpenAI Whisper for accurate transcription
• Real-Time Translation: Instant translation using Google Translate
• Multi-Language Support: 18+ languages including English, Spanish, French, German, Persian, Arabic, Chinese, Japanese, and more
• User-Friendly Interface: Clean Tkinter-based GUI with real-time text display
• macOS Optimized: Compatible with both Intel and Apple Silicon (M1/M2) Macs

📋 Prerequisites

System Requirements

• macOS 10.15 (Catalina) or later
• Python 3.8 or later
• At least 4GB RAM (8GB recommended for better performance)
• Internet connection (for translation service)

Required Software

1. Python 3.8+: Download from python.org
2. Homebrew: Install from brew.sh
3. Xcode Command Line Tools: Run xcode-select --install

🛠️ Installation Guide

Step 1: Clone or Download the Project

# If using git
git clone <repository-url>
cd voice-translator

# Or download and extract the files to a folder

Step 2: Install System Dependencies

# Install Homebrew (if not already installed)
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Install PortAudio for audio processing
brew install portaudio

# Install FFmpeg (required by Whisper)
brew install ffmpeg

Step 3: Set Up Python Environment

# Create a virtual environment
python3 -m venv venv

# Activate the virtual environment
source venv/bin/activate

# Upgrade pip
pip install --upgrade pip

Step 4: Install Python Dependencies

# Install all required packages
pip install -r requirements.txt

# If you encounter issues with torch on Apple Silicon:
pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu

Step 5: Install BlackHole (Optional - for System Audio Capture)

To capture audio from video calls (Zoom, Skype, FaceTime, etc.):

1. Download BlackHole from existential.audio/blackhole
2. Install the .pkg file
3. Configure Audio MIDI Setup:
   • Open "Audio MIDI Setup" (Applications > Utilities)
   • Create a "Multi-Output Device"
   • Select both your speakers and BlackHole
   • Set this as your default output device

🚀 Usage

Running the Application

# Make sure you're in the project directory and virtual environment is activated
source venv/bin/activate

# Run the application
python voice_translator.py

Using the Interface

1. Language Selection:

   • Choose source language (or "Auto Detect")
   • Select target language for translation

2. Audio Input:

   • Microphone: Captures your voice directly
   • System Audio: Captures all system audio (requires BlackHole setup)

3. Controls:

   • Start Listening: Begin audio capture and processing
   • Stop Listening: Stop the process
   • Clear Text: Clear both text areas

4. Real-Time Display:

   • Left panel shows original transcribed text
   • Right panel shows translated text
   • Timestamps are included for each entry

Tips for Best Results

• Audio Quality: Ensure clear audio input with minimal background noise
• Speaking Pace: Speak clearly and at a moderate pace
• Language Detection: Use "Auto Detect" for mixed-language conversations
• System Audio: Make sure BlackHole is properly configured for video call capture

🔧 Configuration Options

Supported Languages

• Source Languages: Auto-detect, English, Spanish, French, German, Italian, Portuguese, Russian, Chinese, Japanese, Korean, Arabic, Persian, Turkish, Dutch, Swedish, Norwegian, Danish, Finnish
• Target Languages: All of the above except Auto-detect

Audio Settings

• Sample Rate: 16kHz (optimized for speech)
• Chunk Duration: 3 seconds (adjustable in code)
• Processing: Real-time with minimal latency

🐛 Troubleshooting

Common Issues

1. "No module named 'whisper'" Error

pip install openai-whisper

2. Audio Device Not Found

• Check that your microphone/audio device is connected
• Grant microphone permissions in System Preferences > Security & Privacy

3. BlackHole Not Working

• Restart the application after installing BlackHole
• Check Audio MIDI Setup configuration
• Ensure BlackHole is set as input device

4. Translation Errors

• Check internet connection
• Try different source/target language combinations
• Restart the application if translation service becomes unresponsive

5. Performance Issues

• Close other resource-intensive applications
• Use "base" Whisper model (default) for better performance
• Consider upgrading RAM if processing is slow

macOS Permissions

The app may request permissions for:

• Microphone Access: Required for audio capture
• Accessibility: May be needed for system audio capture

Grant these permissions in System Preferences > Security & Privacy.

🔄 Updates and Improvements

Potential Enhancements

• Offline Translation: Add support for offline translation models
• Audio Recording: Save audio clips with translations
• Custom Models: Support for specialized Whisper models
• Hotkeys: Global shortcuts for start/stop
• Themes: Dark mode and custom UI themes

Performance Optimization

• Use smaller Whisper models ("tiny", "small") for faster processing
• Adjust chunk duration based on your needs
• Consider using GPU acceleration if available

📝 Technical Details

Architecture

• Audio Capture: sounddevice library with real-time streaming
• Speech Recognition: OpenAI Whisper (local processing)
• Translation: Google Translate API via googletrans
• GUI: Tkinter (cross-platform, included with Python)
• Threading: Separate threads for audio processing and UI updates

File Structure

voice-translator/
├── voice_translator.py    # Main application
├── requirements.txt       # Python dependencies
├── README.md             # This file
└── venv/                 # Virtual environment (created during setup)

🆘 Support

If you encounter issues:

1. Check the troubleshooting section above
2. Ensure all dependencies are properly installed
3. Verify macOS permissions are granted
4. Try running with different language combinations

📄 License

This project is for educational and personal use. Please respect the terms of service for the translation services used.

---

Note: This application requires an internet connection for translation services. Speech recognition (Whisper) works offline once the model is downloaded.