[dannyw]

Because the latent space to picture pipeline is considerably different. There's no weight to output compatibility.

If you ask styleGAN to generate a specific image, that's possible, but you are no longer looking at how well these models generate images.