SEEP

Real-Time AI Face Swap

High-performance real-time face swapping powered by deep learning

---

Features

Feature	Description
Real-Time Processing	Live face swap from webcam with low latency
GPU Accelerated	CUDA/DirectML support for maximum performance
Modern UI	Clean, dark-themed interface built with CustomTkinter
OBS Integration	Virtual camera output for streaming
Face Enhancement	Optional AI-powered face restoration
Multi-Face Support	Swap multiple faces in a single frame

System Requirements

Component	Minimum	Recommended
OS	Windows 10	Windows 11
Python	3.10	3.11
RAM	8 GB	16 GB
GPU	GTX 1060 (6GB)	RTX 3060+
VRAM	4 GB	8 GB+

Note: CPU-only mode is available but significantly slower

Quick Start

1. Clone & Install

git clone https://github.com/mayusi/seep.git
cd seep
pip install -r requirements.txt

2. Download Models

python download_models.py

Or run the app - it will download models automatically on first launch.

3. Run

python main.py

Usage

GUI Mode (Default)

python main.py

Launches the graphical interface with:

• Source face selection panel
• Live preview window
• Camera controls
• Recording options

Command Line Options

python main.py --check      # Check system requirements
python main.py --download   # Download models
python main.py --debug      # Enable debug logging

Building Executable

Create a standalone .exe for distribution:

python build_exe.py

Output will be in dist/Seep.exe

Project Structure

seep/
├── main.py              # Application entry point
├── launcher.py          # GUI launcher with splash screen
├── requirements.txt     # Python dependencies
├── src/
│   ├── ui/              # User interface components
│   ├── core/            # Core processing pipeline
│   ├── face/            # Face detection & analysis
│   ├── capture/         # Camera capture
│   ├── processors/      # Image processors
│   └── utils/           # Utility functions
├── config/              # Configuration files
├── models/              # AI models (downloaded separately)
└── assets/              # Icons and resources

Models

Models are downloaded automatically on first run. They include:

Model	Purpose	Size
inswapper_128.onnx	Face swapping	~530 MB
det_10g.onnx	Face detection	~16 MB
w600k_r50.onnx	Face recognition	~174 MB

Performance Tips

• Use a dedicated GPU for best performance
• Close other GPU-intensive applications
• Lower resolution for higher FPS
• Enable face caching for static source images

License

MIT License - See LICENSE for details.

---

_{Built with passion by Naxte}