Tungsten faceswap

Take a video and replace the face in it with a face of your choice. You only need one image of the desired face. No dataset, no training, no GPUs.

Prerequisites

• Video clips
• Python 3.9
• Docker

Prepare video clips

Copy your video clips on the project root directory.
Then, update tungsten_model.py corresponding to your needs:

1. Change FaceswapInput class - This class is for defining inputs.
2. Change VIDEO_SEGMENTS variable - Change file names to your owns.
3. Change REFERENCE_FRAME_NUMBERS variable - Change frame number containing the target face.

Build, push and run your model

Step 0: Install Requirements

First, install requirements:

pip install -r requirements.txt

Step 1. Prepare weights

Download weights:

./download_weights.sh

Step 2. Build model

tungsten build . faceswap:myversion

Step 3: Create a project on Tungsten

Go to tungsten.run and create a project.

Step 4: Push the model to Tungsten

Log in to Tungsten:

tungsten login

Add tag of the model:

tungsten tag faceswap:myversion <project_name>

Push the model to the project:

tungsten push <project_name>

Step 5: Run the model on Tungsten

Visit tungsten.run and go to the project page.