//=time() ?>
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
i used this older result of mine for the init image: "a monument to the psychedelia of humanity, by Ross Tran and Zdzislaw Beksinski"
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
ruDALL-E for init. image + CLIP guided diffusion
#ruDALLE + #VQGAN + #ClipGuidedDiffusion #MachineLearning #aiart #digitalart #conceptart #fanart #scifiart
It’s Grace’s mans birthday init @solo_vyn ILYSM MORE THAN SOLOMON MARRY ME INSTEAD 💕💕💕 💍
#solomon #obeymesolomon #ObeyMeBD #obeyme #obeymefanart #ソロモン
Starting with the idea of doing zodiac/astro signs, here I got messed up with Scorpio init image and Pisces prompt, the result is better than I could have thought up
@Pixelmindai Source image - Result
(Query: .diffusion Portrait of Neo Tokyo Citadel Ratchitect, Race - Rat, Eyes - Futuristic Goggles, Facial hair - short beard, Class - Architect, Wavy Hair, Gender - Male. Intelligent. Trending on artstation -skip_timesteps 65 -init_scale 1000)
Day 10 | Senku Ishigami ( Dr. Stone )
I love this show and the characters init.
#arttrober2021
Step one: draw an init image to guide rough composition.
Feed it into VQGAN+CLIP with a suitable prompt - I tried "A spooky carnival at night, torches burning as the wheel of time turns ominously. Oil Painting by Sci-Fi concept artist Mr Fire"
Marvel at the results and pick one
That one on the right side is the 512 CLIP Guided Diffusion model.
The same init and settings on the 256 guided diffusion model stray far from the pixelated look, I assume because the init had to be stretched/upsampled more for the 512