DeepFaceLive

Real-time face swapping and animation for video calls and streaming

2024-05-07

DeepFaceLive is an advanced AI-powered tool designed for real-time face swapping and animation. It allows users to seamlessly replace faces in live webcam feeds or pre-recorded videos using either pre-trained models (publicly available or custom-trained with DeepFaceLab) or a single photo. The application also features a Face Animator module, enabling control of static face images via live video input, ideal for creating entertaining content or streaming at 25 fps with high-end GPUs.

Key features include:

  • High-quality face swapping: Supports both public models (like the included Keanu Reeves model) and custom-trained models for better personalization.
  • Real-time performance: Optimized for modern GPUs (recommended RTX 2070+ or Radeon RX 5700 XT+), achieving smooth operation during video calls or streaming.
  • Portability: Comes as a zero-dependency, self-extracting folder that works without installation (except for video drivers).
  • Multi-platform support: Available for Windows and Linux (with build instructions).
  • DirectX12 and NVIDIA-specific builds: For broader compatibility or optimized NVIDIA performance.

DeepFaceLive is perfect for content creators, streamers, or anyone looking to experiment with AI-driven face manipulation in real-time. The project encourages community contributions through model sharing via Discord and provides detailed FAQs for training custom models. Note: Public models are of fictional characters, with similarities to real people being coincidental (except Keanu Reeves, who is explicitly included).

Artificial Intelligence Face Swapping Real-time Processing Deep Learning Video Editing