//=time() ?>
Four DIP + CLIP ViT-B/32 outputs for "a gateway between dreams", using a new model I am training to sample CLIP image embeds given a text embed (like the DALL-E 2 prior). It considerably increases the diversity of outputs for a prompt for things like DIP+CLIP and VQGAN+CLIP!
"a beautiful epic wondrous fantasy painting of the alphabet" Deep Image Prior + CLIP ViT-L/14 336px
"the oracle at Delphi by Gerardo Dottori"
"the oracle of waters by Karel Thole"
Deep Image Prior + CLIP ViT-L/14 336px
"a vase of flowers in a scrambled, disjointed style" #midjourney diffusion v2
"liberation of the emotions, in a symbolic, meaningful, and vibrant style" #midjourney diffusion v2
"liberation of the mind, in a symbolic, meaningful, and vibrant style" #midjourney diffusion v2
"Resurrection of the dead via multiverse-wide acausal cooperation[, in a symbolic and meaningful style]" #midjourney diffusion v2