Free AI face-swapping solution - VisoMaster-Fusion Batch Video and Image Face Swap

Today I'm sharing some tools suitable for face swapping videos/images.

When I want to create funny videos or pictures, I always end up with poor results when using Photoshop for face swapping. AI But it's a bit expensive.

I finally found something great on GitHub called VisoMaster-Fusion, which can perform face-swapping on videos or images locally, and my 3060 can handle it.

VisoMaster Fusion is a Windows desktop application. AI Face processing software is mainly used for face replacement, face restoration, expression editing, mask control, and batch processing in images, videos, and real-time camera footage.

Project address:https://github.com/VisoMasterFusion/VisoMaster-Fusion

VisoMaster Fusion supports multiple input sources:

• Images: Common formats such as JPG, PNG, and WebP
• Video: Dependence FFmpegIt supports common video containers such as MP4.
• Camera: Capable of handling real-time webcam input
• Virtual camera: Suitable for live streaming, meetings, and real-time preview scenarios.
• VR180 assets: The project has built-in dedicated VR180 processing support.

VisoMaster Fusion supports multi-face scenarios, allowing you to assign different source faces to different target faces. It also allows you to use a similarity threshold to control which source face replaces only target faces with sufficiently high similarity. This feature is crucial when multiple people appear simultaneously in a video.

When performing face swapping in videos, VisoMaster Fusion also supports ByteTrack tracking, which not only detects each frame but also tries to maintain consistency between consecutive frames.

After the face swap, the project can continue to refine the results.

• Face Restorer: Repairing Facial Details
• Second Restorer Pass: Second Round of Repairs
• GFPGAN-1024: Face Enhancement
• ReF-LDM Denoiser: Noise Reduction and Texture Reconstruction
• Frame Enhancer: Enhances or amplifies the entire frame.
• AutoColor Transfer:automaticColor migration
• Ending Color Transfer: Final Color Unification
• Texture Transfer: Transferring textures from the original face
• JPEG/MPEG artifact simulation: Simulates compressed textures to make the results closer to the video source.

These features are not just for beautification; they are more about ensuring that the clarity, noise, color, and compression artifacts of the face-swapped area remain consistent with the original video.

VisoMaster Fusion also offers a Face Editor and expression-related modules to address dynamic issues in videos. For example, when a target person is speaking, turning their head, looking down, or opening their mouth, ordinary face swapping can easily result in stiff expressions or misaligned mouths.

The official documentation says that 6GB of NVIDIA VRAM is enough to run it, and my 3060 12G runs it very smoothly. However, the installation package requires 20GB to 30GB of disk space, so remember to reserve some.

If you clone the repository yourself, it requires several steps:

git clone https://github.com/VisoMasterFusion/VisoMaster-Fusion cd VisoMaster-Fusion

2. Create and activate the Python environment

conda create -n visomaster python=3.12 -y conda activate visomaster pip install uv

3. Install dependencies

uv pip install -r requirements_cu13.txt

4. Download the required model

python download_models.py

5. Install FFmpeg

winget install -e --id Gyan.FFmpeg --version 7.1.1

After downloading and extracting, add the following directory to your Windows PATH:

<解压后的 ffmpeg 路径>\bin

6. Double-click to run

My suggestion is to just let Claude Code or Codex ThisAgent It's very easy to install.

It may seem time-consuming, but it's very convenient for users unfamiliar with command-line usage. If problems arise, it's easy to get help... Agent Codex helped me fix it and even reminded me to upgrade my driver.

Case 1: Image face swapping

After opening the software, select the image or video to be face-swapped in the Target Media section in the upper left corner. The folder will then open.automaticDisplay video and image thumbnails.

In the middle left section, select "Input Faces" to provide a photo of your face; it's best to use a clear, front-facing photo.

We select the object whose face we want to swap in Target Media, and click Find Face in the bottom middle. Target Media will recognize the face in the target material, and the recognized face will be displayed in the list below.

In Input Faces, select the source face, then in the target face list, select the face you want to replace, and then click the Swap Faces preview area to generate the face-swapping result.

If you find the generated image blurry, you can change the Swapper Resolution under Swapper in Face Swap on the right from 128 to 256 or 512.

If it's still a bit blurry, tap Restorers on the right to enable Face Restorer. First, select GFPGAN-v1.4 for Restorer Type. If it's still blurry, try GFPGAN-1024.

To save, go to Settings on the right, find Output Directory, click the folder button next to it, and select the save directory.

Click "Save Image" in the bottom left corner to save the current image.

Admire Ultraman's stunning good looks:

Case 2: Video face swapping

The same steps apply to videos. Select the video you want to have your face swapped in Target Media, click Find Face in the bottom middle, and Target Media will recognize the faces in the target material. The recognized faces will be displayed in the list below.

In Input Faces, select the source face, then in the target face list, select the face you want to replace, and then click the Swap Faces preview area to generate the face-swapping result.

Click the record button (the circle icon indicating the video save point). The software will start processing the video from the current time point and save the result to the directory you just set.

Let's take a look at the saved video:

AI Video content creation is no longer something only a few teams can handle, and it's becoming increasingly lightweight.

According to the Wyzowl 2026 report, 91% of surveyed companies are already using video for marketing, and 63% of video marketers have used it. AI Video tools.

The value of VisoMaster Fusion lies in the fact that when you have a lot of materials, make frequent modifications, and need to preview repeatedly, running it locally is more convenient than uploading it to an online platform each time.

By putting face detection, replacement, repair, masking, batch processing, and video export into the same interface, we don't have to switch tools every step of the way.

VisoMaster Fusion is suitable for processing licensed assets, entertainment creation, special effects testing, and virtual content.Please do not use real people's images to create misleading, impersonating, or privacy-invading content..

Original link:A gem of a low-starred GitHub project: a face-swapping project that can run locally on a 3060-bit system.