@vgan @maltimore I agree. I think you can see that when you prompt these networks to output faces. I couldn't get vqgan+clip to output a face that didn't look very distorted but the latent diffusion model im currently using gives decent results at least for some celebrity name in the prompt
#vqgan #clip #LatentDiffusion #text2art
#vqgan #clip #LatentDiffusion #text2art