This repo contains two jupyter notebooks that provide a minimal example for how one may interact with the official Stylegan2 TensorFlow implementation.

The work in this repo relies on and was based heavily on the following repos:

• https://github.com/NVlabs/stylegan2
• https://github.com/rolux/stylegan2encoder
• https://github.com/Puzer/stylegan-encoder
• https://github.com/a312863063/generators-with-stylegan2

The jupyter notebooks include code that allows one to:

• Convert any JPG or PNG image into their StyleGan2 latent representations
• Modify the latent representations
• Generate images from latent representations

Several precomputed "latent directions" are also made available. These latent directions are pulled from the following repo:

https://github.com/a312863063/generators-with-stylegan2

For example, This allows one to move a latent vector in the smile direction to increase an image's smile.

For a complete list of latent directions available, see:

https://github.com/a312863063/generators-with-stylegan2/tree/master/latent_directions