StyleGAN3 Remake: Who The Fuck Is Conor?

As my first deeper exploration of using Machine Learning to create images, I remixed my previous film Who The Fuck Is Conor? with a GAN.

I used ffmpeg to extract a few thousand frames from the film, and finetuned a StyleGAN3 model on these frames. I trained a smaller model on the opening animation sequence in order to retain that.

Next I performed a search in the latent space of the trained model for a frame from each second, and interpolated between them.

Here are some outputs the outputs from the models latent space:

A grid of images produced from a GAN AI model.
A grid of images produced from a GAN AI model.