Deepfake Live Camera for macOS M2

Real-time face swap application optimized for Apple Silicon (M1/M2/M3) with OBS Studio integration.

Developed by Ilyas AKKUS

Features

• Real-time face swapping using a single source image
• Optimized for Apple Silicon using CoreML
• OBS Studio virtual camera integration
• Low latency for live streaming
• Simple GUI interface

Requirements

• macOS with Apple Silicon (M1/M2/M3)
• Python 3.10 (IMPORTANT: Must use Python 3.10, not newer versions)
• OBS Studio
• Webcam

Installation

1. Install Python 3.10

# Install Python 3.10 using Homebrew
brew install python@3.10

# Verify installation
python3.10 --version

2. Create Virtual Environment

# Create virtual environment with Python 3.10
python3.10 -m venv venv

# Activate virtual environment
source venv/bin/activate

3. Install Dependencies

# Install requirements
pip install -r requirements.txt

4. Download Models

Models will be automatically downloaded on first run (~300MB).

Usage

Basic Usage

1. Activate virtual environment:

   source venv/bin/activate

2. Run the application:

   python3.10 run.py --execution-provider coreml

3. Select source face image

4. Click "Live" to start webcam mode

5. Use OBS Studio to capture the output window

OBS Studio Integration

1. Open OBS Studio
2. Add a new "Window Capture" source
3. Select the Deepfake Live Camera preview window
4. Alternatively, use "Display Capture" to capture the entire screen

Tips for Best Results

• Use high-quality source images with clear face visibility
• Ensure good lighting for webcam
• Source image should have similar angle to your webcam position
• Close-up portraits work best

Troubleshooting

Common Issues

1. "_tkinter not found" error:

   brew reinstall python-tk@3.10

2. Python version conflicts:

   # Always use python3.10 explicitly
   python3.10 run.py --execution-provider coreml

3. Performance issues:

   • Make sure to use CoreML execution provider for Apple Silicon
   • Close unnecessary applications
   • Reduce preview window size if needed

M2 Optimizations and Performance

Performance Modes

The application supports three performance modes that can be set using the --performance flag:

python3.10 run.py --performance [fast|balanced|quality]

• fast: Best performance, lower quality

  • Reduced resolution (40% of original)
  • Higher frame skip rate
  • Recommended for live streaming

• balanced: Good balance between performance and quality

  • Medium resolution (50% of original)
  • Moderate frame skip rate
  • Good for general use

• quality: Highest quality, lower performance

  • Higher resolution (80% of original)
  • Lower frame skip rate
  • Best for recording

M2-Specific Optimizations

1. CoreML Settings:

   • Uses Neural Engine for inference
   • Optimized model conversion for M2
   • Efficient memory management

2. System Recommendations:

   • Set Mac power settings to "High Performance"
   • Ensure proper cooling
   • Close resource-intensive applications
   • Keep macOS updated

3. Performance Tips:

   • Use "fast" mode for live streaming
   • Disable lip sync if not needed
   • Use lower resolution webcam if available
   • Monitor FPS counter in the application

Advanced Settings

You can combine performance settings with other options:

# Fast mode with specific camera
python3.10 run.py --performance fast --camera-id 1

# Balanced mode without GUI
python3.10 run.py --performance balanced --no-gui --source path/to/image.jpg

License

This project is licensed under AGPL-3.0 License.