SimSwap

What it does

Blending options

It supports mask-based and bounding-box methods for placing the changed face back into the original image, and recommends using a mask.checked 11 Oct 2026

Dependencies

The README lists Python 3.6+, PyTorch 1.5+, torchvision, OpenCV, Pillow, NumPy, imageio, moviepy, InsightFace, and timm 0.5.4 as dependencies.checked 10 Oct 2026

Face selection

It supports swapping the detected face with the highest confidence, all faces, or specified faces, including multiple specified faces.checked 11 Oct 2026

Hardware requirement

The maker expects users to have a GPU with at least 3 GB of memory and points users without one to a Colab notebook implementation.checked 11 Oct 2026

High resolution

The usage guide describes an open-sourced 512-pixel beta model trained on VGGFace2-HQ.checked 11 Oct 2026

Image and video modes

The usage guide documents face swapping in both images and videos.checked 11 Oct 2026

Image support

It supports swapping faces in aligned images and arbitrary images, including replacing one face or all faces in an image.checked 10 Oct 2026

Installation

The preparation guide recommends Anaconda and lists PyTorch, torchvision, InsightFace, ONNX Runtime, and MoviePy among the setup dependencies.checked 11 Oct 2026

Integrations

The README links to a Colab demo and an image-swapping demo and Docker image on Replicate.checked 10 Oct 2026

Intended use and license

The repository says the project is for technical and academic use and is licensed for academic and non-commercial use under CC-BY-NC 4.0.checked 11 Oct 2026

License and use

The project states it is for academic and non-commercial use only and is released under CC-BY-NC 4.0.checked 10 Oct 2026

Masking

The usage guide recommends using a mask for better visual effects and also describes bounding-box pasting as an alternative.checked 10 Oct 2026

Operating systems

The maker says its recommended setup has been tested on Windows and Ubuntu.checked 10 Oct 2026

Purpose

SimSwap is a face-swapping framework for images and videos that uses one trained model.checked 11 Oct 2026

Requirements

The preparation guide recommends a GPU with at least 3 GB of memory and points users without one to a Colab notebook implementation.checked 10 Oct 2026

Support

The maker invites users to ask questions in GitHub issues or contact the listed project email addresses.checked 11 Oct 2026

Supported operating systems

The maker says the documented setup has been tested on Windows and Ubuntu.checked 11 Oct 2026

Use guidance

The README says the project is for technical and academic use and asks users not to apply it to illegal or unethical scenarios.checked 10 Oct 2026

Video support

Video options include swapping the highest-confidence face, all faces, a specified face, or multiple specified faces.checked 10 Oct 2026

Watermark

The example commands add the SimSwap logo watermark by default, and the `--no_simswaplogo` option controls its removal.checked 10 Oct 2026