DragGAN is a project developed by Google researchers together with the Max Planck Institute of Informatics and MIT CSAIL. It is a very intuitive image editing tool. Users only need to control the pixels and directions in the image to quickly adjust The position, posture, expression, size and angle of the subject of the photo.

Download pretrained StyleGAN2 weights

To download pretrained weights, just run:

sh scripts/download_model.sh

Run DragGAN GUI

To start the DragGAN GUI, simply run:

This GUI supports editing images generated by GANs. To edit a real image, one needs to first perform a GAN inversion using a tool such as PTI. The new latent code and model weights are then loaded into the GUI.