A Roman mosaic of the Teenage Mutant Ninja Turtles
(Not sure who didin’t make the cut)
Photorealistic Scooby Doo eating ice cream in Bruges
(Definitely not what I had in mind)
A Roman mosaic of the Teenage Mutant Ninja Turtles
(Not sure who didin’t make the cut)
Photorealistic Scooby Doo eating ice cream in Bruges
(Definitely not what I had in mind)
4 Likes
I couldn’t be bothered jumping through all the hoops to get Stable Diffusion running on a Radeon card so I played around with the DreamStudio beta for a bit, trying out different prompts for a realistic picture of Hannibal crossing the Alps. They throttle this by having each generation take progressively longer so I gave up after about 15 minutes. But uh, yeah. 15 minutes of fiddling. On beta software. And it’s open source. Lawnmower_Man started this thread less than a year ago. Scroll up and see what the cutting edge was last November.
3 Likes
If only he were wearing gloves and chomping a cigar…
4 Likes
ok this stuff is legit amazing
1 Like
Jfc I had no idea there was a Version 4 and that I was feeding the bot outdated prompts.
wes anderson’s ALIEN
2 Likes
midjourney v4 is a huge leap over v3 and is turning out some insane shit
I feel like it’s very sensitive to the inputs. Most of the amazing ones I’ve seen are drawing from high quality cinematography stills. In terms of just base v4, it’s certainly better than base v3 but not crazy better than --testp --creative flags I was running on everything for last few months. The big advantage over those flags though is that you can feed it image prompts and also reprompt on the variations. It’s still off a lot of times on capturing certain likenesses and also hands / arms.
How did they create the images for Jodorowsky’s Tron?
Still blown away by those.
I found it:
Darrell also said, “My prompts were simple variations of “production still from 1976 of Alejandro Jodorowsky’s TRON –ar 3:2” and then added in extra details like film lenses and “light cycle” or “disk” or “Recognizer” etc.“
1 Like