A face-swap app that actually ships with tests and a conscience

What it does

VisoMaster Fusion is a Windows desktop application for AI-powered face swapping and enhancement across images, videos, and live webcam feeds. It bundles multiple swapper models (Inswapper128, SimSwap, GhostFace, DeepFaceLive DFM, and others) behind a graphical workflow with timeline markers, batch job management, and output controls for multi-segment renders.

The interesting bit

The project is essentially a polished integration layer — it combines the original VisoMaster work with community mods for VR180, a Job Manager, and a ReF-LDM denoiser, then wraps the whole thing in a portable launcher that downloads Python 3.12, PyTorch, CUDA, TensorRT, and model files on first run. The maintainers also bolted on a test suite covering VR math, face masks, serialization, and job validation that runs headless without Qt or a GPU, which is almost unheard-of in this corner of open-source.

Key highlights

• Portable launcher with self-update, repair, dependency check, and model optimization tools
• Multi-face workflows with source face cards, saved embeddings, similarity thresholding, and optional ByteTrack tracking
• VR180 support and ReF-LDM denoiser with single-step and DDIM modes
• Job Manager for saving workspaces and running batches unattended
• TensorRT default provider with CUDA and CPU fallbacks
• Eleven built-in themes including OLED-Black, Nord, and Gruvbox

Caveats

• Windows 10/11 only; Nvidia GPU strongly recommended (6 GB VRAM minimum, 8–12 GB for heavier work)
• Requires internet on first run and 20–30 GB disk space
• CPU inference works but is “much slower” per the README

Verdict

Worth a look if you need a packaged, feature-heavy face-swap workflow on Windows and don’t want to hand-roll your own model pipeline. Skip it if you’re on Linux, GPU-poor, or allergic to binary downloads that phone home on first launch.