DeepFaceLive: Real-Time Face Swapping for Video

Overview

DeepFaceLive is a Python application for real-time face swapping in PC streams and video calls. It can use a trained face model or, for its Insight swap mode, a single photo. A separate Face Animator module lets video or a camera feed control a still face image.

It is suited to people who want live, on-device face effects rather than a post-production workflow. The project is archived, so treat it as a fixed snapshot and check its documentation and requirements before relying on it.

Key Features

• Swap a face from a webcam or video using trained DFM face models.
• Use the Insight face-swap mode with a single photo.
• Animate a static face image using video or a camera feed.
• Provides Windows setup documentation for streaming and video calls.
• Supports Android phone camera input, according to the project documentation.
• Offers DirectX 12 and NVIDIA builds for Windows.
• Includes Linux build information.

Use Cases

• Streamers can add a live face swap to a PC broadcast when they have compatible hardware and a suitable model.
• Video-call users can route a webcam-based face swap into their call setup.
• Creators can apply a face swap to video input without limiting the effect to a prerecorded editing workflow.
• Users making memes or short clips can experiment with controlling a static portrait through the Face Animator module.

Getting Started

Review the Windows main setup guide and the streaming or video-call guide as needed. The README links to standalone Windows releases and additional setup instructions; consult it for the full process and requirements.

Alternatives

• deepface: DeepFace provides face recognition and attribute analysis, while DeepFaceLive swaps faces in real time or animates a still image.
• MuseTalk: MuseTalk synchronizes a face to speech for video dubbing, while DeepFaceLive focuses on live face swapping and still-image animation.

Considerations

• The documented primary setup is Windows 10. The listed requirements include a DirectX 12-compatible graphics card, a modern CPU with AVX instructions, 4 GB RAM, and a 32 GB or larger paging file. RTX 2070+ or Radeon RX 5700 XT+ is recommended.
• Real-time processing depends on capable graphics hardware. Face Animator quality is described as limited and may require careful matching and parameter tuning.
• The repository is archived, and its latest recorded push was 2024-11-08. Do not assume ongoing maintenance or compatibility with newer systems.
• The project is licensed under GPL-3.0. Review the license and obtain consent for face imagery and models before use, especially in public or deceptive contexts.
• See the README and its linked documentation for current release and setup details.