A GUI that puts deepfake tooling within accidental reach

What it does VisoMaster is a desktop application for AI-driven face swapping and editing in images and videos. It bundles multiple swapper models, supports DeepFaceLab-trained models, and adds an expression editor built on LivePortrait. The interface targets casual users rather than command-line wranglers.

The interesting bit The project treats real-time performance as a first-class feature: live playback before export, TensorRT acceleration, and a virtual-camera mode that pipes swapped faces straight into Zoom or Twitch. That is a lot of moving parts to keep in sync, and the authors seem aware of it.

Key highlights

• Multi-face swapping with per-facial-part masking and DFL XSeg occlusion handling
• Expression restorer that carries over original expressions to the swapped face
• Face restoration via “all popular upscaling & enhancement models” (specific models not listed)
• LivePortrait-based pose and color adjustments for face, hair, eyebrows, and lips
• Per-frame video markers for fine-grained control over a sequence
• Windows portable installer with automatic CUDA setup; Linux runs via python main.py

Caveats

• Nvidia GPU required; no mention of AMD, Apple Silicon, or CPU-only paths
• Model downloads and dependency placement are manual steps even on Windows
• The README is upfront that this is a hobby project, and the liability disclaimer is copied verbatim from another repo

Verdict Worth a look if you need a graphical front-end for face-swapping workflows and have the hardware to match. Skip it if you expected a polished commercial product or cross-platform support beyond Windows and Linux.