- generate a bunch of content from a diffusion model.
- wire up a styleGAN-XL for LoRA training
- apply projected GAN, using new FID results/ensembles
- experiment with LoRA/GLoRA + GAN
- experiment with new FID stuff + GAN
- mix n match
- hit up that GAN wizard, see if he's interested or has already done some of this already