"A photo of two 1980s ladies aged around 25. The background is a gameshow with laughing audience. Elise is a redhead with freckles, sitting on a stool looking nervous, wearing a summer dress. Bridget is blonde, sitting on a stool inside a tall glass box with her hands in her lap, wearing a colourful tracksuit."
(It can take a few goes because it sometimes does a split screen rather than a single shot with both contestants).
But of course we can't let Bridget go home clean, can we?
"A photo of two 1980s ladies aged around 25. The background is a gameshow with laughing audience. Elise is a redhead with freckles, sitting on a stool looking nervous, wearing a summer dress. Bridget is blonde, sitting on a stool inside a tall glass box with her hands in her lap, wearing a colourful tracksuit, coated in multicoloured slime. Slime pours out of a pipe in the box over Bridget's head. The slime is dripping off Bridget's head, hands and clothes."
Unfortunately I haven't found a prompt to do that -- when I try to add a tank of slime, whatever paraphrase I use, it insists on putting the tank under Bridget's seat rather than above her head!
Getting them both in the tanks seems to be harder, but something like "A photo of two 1990s ladettes aged around 25, both sitting on stools inside tall glass boxes. The background is a gameshow with laughing audience. Elise is a redhead with freckles, laughing, wearing a summer dress. Bridget is blonde, shocked, with her hands in her lap, wearing a colourful tracksuit, coated in multicoloured slime. Slime pours out of a pipe in the box over Bridget's head. The slime is dripping off Bridget's head, hands and clothes." seems to work. At least sometimes.
(Oh, and I did one where Elise got the gunking. Since she'd been keeping out of the gunge all the way down the thread)
Change the character descriptions around a bit, and you get the scene from the romance novel "Dating Her Boss" (mentioned by maxoverdrive on this thread: https://umd.net/forums/wam-fiction-on-amazon#postid_486533 ) where the heroine loses a messy game show. (She suspects the host's bitchy PA of rigging it, particularly when she deliberately didn't warn her to bring a change of clothes).
Hi there, I have been making my own similar images, I am having difficulty describing the gunge container to Ai as well as getting it to include the nozzle or whatever the gunge dispenser is called, do you have any suggestions. Thanks.
MrGYOB_Ai said: Hi there, I have been making my own similar images, I am having difficulty describing the gunge container to Ai as well as getting it to include the nozzle or whatever the gunge dispenser is called, do you have any suggestions. Thanks.
MrGYOB_Ai said: Hi there, I have been making my own similar images, I am having difficulty describing the gunge container to Ai as well as getting it to include the nozzle or whatever the gunge dispenser is called, do you have any suggestions. Thanks.
I use
"sitting on stools inside identical tall glass boxes with a pipe at the top"
then whatever your substance is, say it pours from the pipe
I have to say, the most exciting thing about this to me is the level of "contextual awareness" DALL-E3 seems to have. The images in this thread are really good, but I find the "tone" of DALL-E to lack a lot of the subtle artistic nuance of MJ or the best SD checkpoints. With those others though, there's just no way to prompt for "X doing Y to Z". That level of interaction between scene elements is beyond them.
Hopefully whatever they did differently in their training (not sure if it's just more params or something a bit smarter) will propagate to the open source model community soon, because everything else is way better in OS - having LoRAs and ControlNets and custom checkpoints makes everything way more exciting, especially for those of us keen to explore the weird specifics of our kink! Can't get over that experience of going through 10s of images that are just "meh" and then suddenly it rolls one that just hits dead-on!
It has been a while since this was posted but hey... I just ran some of my favorite by Aceman through the Krea enhance to make them look less like cartoons. Seems to have worked pretty well.