hacksider/Deep-Live-Cam

What it solves

Deep-Live-Cam enables real-time face swapping and video deepfaking using only a single source image of the target face. It allows users to replace faces in live webcam streams, existing videos, or images instantly.

How it works

The software uses pre-trained AI models (such as GFPGAN and inswapper) to map a source face onto a target subject. It can operate in two primary modes: Image/Video mode for processing files and Webcam mode for live streaming. To ensure performance across different hardware, it supports multiple execution providers including NVIDIA CUDA, Apple CoreML, Intel OpenVINO, and DirectML.

Who it’s for

It is designed for AI-generated media artists, content creators, and performers who need to animate custom characters or create engaging visual content in real-time.

Highlights

• Single-Image Swap: Requires only one photo to perform a face swap.
• Real-Time Webcam Integration: Supports live face swapping for video calls or performances.
• Mouth Masking: Ability to retain the original mouth for more accurate movement.
• Face Mapping: Supports applying different faces to multiple subjects simultaneously.
• Broad Hardware Support: Optimized for Windows, Mac Silicon, and various GPUs (NVIDIA, AMD, Intel).