How to Use Faceswap AI?


Faceswap AI is an open-source deepfake tool that allows you to swap faces in videos. It uses deep learning techniques to automatically replace one face with another with a high degree of realism. In this guide, we will walk through the entire process of using Faceswap to create deepfake videos, from installing the software to swapping faces in videos.

Recent Released:How to Use Murf AI?

Overview of Faceswap

Faceswap is a free, open-source deepfake software that works on Windows, Linux and macOS. It uses deep learning to automatically replace the face of one person in a video with someone else’s face. Some key features of Faceswap include:

  • Automatic face detection and swapping in videos
  • Support for swapping faces between photos and videos
  • Options for training your own models or using pre-trained models
  • GUI and command line usage
  • Detailed documentation and active community forum

While deepfakes can be used maliciously, Faceswap enables users to safely learn and experiment with AI and deep learning. When used ethically, it has many creative applications, like creating fun videos for friends.

Step 1 – Install Faceswap

The first step is to install Faceswap on your computer. Here are the installation instructions for different operating systems:


  1. Download the latest version of Faceswap from the official website. Choose the Windows package.
  2. Unzip the downloaded file and copy the Faceswap folder to your desired installation location. For example, C:\Faceswap.
  3. Open the Faceswap folder and launch run.bat. This will install the required Python dependencies.


  1. Install the required dependencies: gitpython3python3-devpython3-venvlibsm6libxext6 and libxrender-dev.
  2. Clone the Faceswap repo: git clone
  3. Enter the faceswap folder: cd faceswap
  4. Create and activate a virtual environment: python3 -m venv venv && . venv/bin/activate
  5. Install Faceswap’s python dependencies: pip3 install -r requirements.txt


  1. Install brew if you don’t already have it: /usr/bin/ruby -e "$(curl -fsSL"
  2. Install the required dependencies: brew install git python3 python3-pip
  3. Clone the Faceswap repo: git clone
  4. Enter the faceswap folder: cd faceswap
  5. Create and activate a virtual environment: python3 -m venv venv && . venv/bin/activate
  6. Install Faceswap’s python dependencies: pip3 install -r requirements.txt

Once Faceswap is installed, you can launch it by running python from the repo folder. This will open the Faceswap GUI.

Step 2 – Extract Faces from Videos

Before swapping faces, you need to extract frames containing faces from the source videos. Faceswap provides a tool to automate this.

To extract faces:

  1. Open the Faceswap GUI.
  2. Go to the Extract tab.
  3. Click Load and select a video file containing the face you want to extract. Multiple video files can be added.
  4. Set the output folder where extracted faces will be saved.
  5. Optional – Adjust the extract settings like filter, detector, and aligner. The defaults are fine for most videos.
  6. Click Extract to start extracting faces.

This process extracts faces from the source videos and saves them as images in the output folder. Multiple faces will be extracted from each video. These extracted faces will be used for training the face swap model.

You’ll need to repeat this process for all the videos containing faces you want to swap. Extract both the face you want to replace and the face you want to swap in.

Show Image

Step 3 – Choose and Train a Face Swap Model

Once you’ve extracted the required faces, the next step is choosing and training a model to swap the faces.

Faceswap offers several different model architectures. The two main ones are:

  • Original: The standard model that provides a good balance of quality and speed.
  • Lightweight: A smaller and faster model, but can have lower image quality. Good for quick previewing.

For swapping faces between photos and videos, the original model is recommended.

To train the model:

  1. Go to the Train tab in Faceswap’s GUI.
  2. Pick the model type – original or lightweight.
  3. Set the path to the extracted faces for the person you want to replace as Face A.
  4. Set the path to the extracted faces for the person you want to swap in as Face B.
  5. Set a folder path for the output of the training.
  6. Adjust any other model settings you want to tweak. The defaults are generally good.
  7. Click Train to start the training process.

Training can take several hours to multiple days depending on your GPU, the amount of training data, and the model architecture. Faceswap leverages your GPU for faster training.

The training process will iterate through the extracted faces and generate a face swap model tailored to the faces you provided.

Show Image

Step 4 – Convert Videos with the Trained Model

Once training completes, it’s time to use the trained model to swap faces in your videos.

To convert a video:

  1. Go to the Convert tab in Faceswap’s GUI.
  2. Pick the trained model file. This is in the models folder inside the output directory set during training.
  3. Set the input video as the Input Video which contains the face you want to replace.
  4. Set the output video path as Converted Video where the face swapped video will be generated.
  5. Adjust any conversion settings if desired.
  6. Click Convert to start swapping faces in the input video using the trained model.

The conversion process runs frame-by-frame through the input video, detects faces, and swaps them with the face from Face B using the trained model. This can take a while depending on the input video length and resolution.

The end result will be a new video with the desired face swapped realistically. You can repeat this process for other videos using the same trained model.

Show Image

Advanced Usage

The guide above covers the core face swapping workflow in Faceswap. Here are some advanced tips and options you can explore:

  • Extract more varied high quality faces for better training results – different angles, lighting, expressions, etc.
  • Experiment with the model architecture, training iterations, learning rate and other hyperparameters.
  • Use mask blurring for seamless face blending.
  • Swap multiple faces simultaneously.
  • Train personalized models focused on specific face pairs for optimal quality.
  • Use the command line interface for advanced workflows and automation.

Be sure to check the official Faceswap forum for detailed guides on all aspects of the tool. The community is also very helpful for troubleshooting issues.


This concludes our step-by-step guide on using Faceswap for deepfakes. To recap:

  • Faceswap is a powerful open source tool for swapping faces using deep learning.
  • The key steps are installing Faceswap, extracting faces from videos, training a model on the faces, and swapping faces in videos with the model.
  • With the right training data and model tuning, Faceswap can produce highly realistic face swaps.
  • There are many advanced customization and automation options available via the CLI and Python API.

We’ve only scratched the surface of what’s possible with AI face swapping. As the technology improves, deepfakes will become more accessible and easier to create. However do note that deepfakes can be used to spread misinformation or violate someone’s privacy. As with any technology, ensure you use Faceswap responsibly and ethically.

Leave a Comment

%d bloggers like this: