Got Milk? (AP) (AI)

[BBB]

In 1984, the National Dairy Board launched “Milk: It Does A Body Good,” a campaign designed to promote dairy milk as the key to strong bones, healthy joints, and an attractive body. In the commercials, young boys and girls chat up their crushes while drinking milk and slowly transforming into fit and beautiful adults.

(from https://sentientmedia.org/milk-ad-history/)

Here is a continuation of that theme, with a girl dutifully drinking her milk at breakfast.
https://www.imagebam.com/view/GAG8NK

Despite using AI, this turned into quite a project. I've taken a ridiculous amount of time and generated a ridiculous number of images to test what effect every word in the prompt has. I had to define an art style I liked, figure out how to lay out the characters, how to specify details of one version without affecting the others, identify the characteristics that would make them appear to be different ages while remaining the same person, and line it up in a before-during-and-after sequence that follows a logical flow of time. But... I feel like I'm getting better results than my last foray.

One of the hardest parts is getting the clothes to burst off. It does makes sense, since a tiny proportion of clothes bursting images exist compared to the huge number of 'clothes on' and even the number of 'clothes off' images out there on the internet for the AI to train on.

My process is to save the generated images that have a kernel of 'good' art (almost a thousand). Then whittle it down to the ones that have one or two flaws, that I'm able and willing to fix in Photoshop. The most common flaws have been:

• Changing the color/pattern of the character's hair or clothes
• Changing the style of clothes, like short pants to long pants
• Clothing in illogical places, like a shirt on the ground when she's still wearing a shirt.
• Extra text - labels, onomatopoeias, etc. I went with "Got Milk" instead of "Milk: It Does a Body Good" because the AI is less likely to screw up the text of a shorter slogan.

You can try to lock down every little detail in the prompt. But the longer your prompt gets, the greater the risk that the AI ignores some details. Plus I wanted to explore a creative variety, instead of the exact same details every time.

This is the first batch I finished, but I have more saved which I intend to finish and share in time.

[hiper]

Is there a way to get the deleted imagebam.com continuation back?

[BBB]

Criminy. I don't know why the imagebam gallery was deleted. The earlier one I did, which as it happens had more nudity, is still up. Well, the good news is that wasn't a continuation. It was a just copy of this set of 20 uploaded here, which I made for redundancy and/or the benefit of non-registered users. I suppose I'll re-upload it, and hopefully have a batch of new images soon.

[BBB]

A second batch of girls drink their milk.

https://www.imagebam.com/view/GAGDUH

[BBB]

Imagebam won't let me re-upload the first batch. Every time I try, it says "Your images have been uploaded!" but shows zero images and no link to the gallery. I deduced what the link actually is, but it shows no images, unlike the second batch of new images which show up right away.

Nonworking gallery of batch 1: https://www.imagebam.com/view/GAGDUT

I tried on two different browsers, with different session IDs. It's as if it remembers those particular images and prevents them from being shown ever again. In the past, imagebam just worked with no issues. As a result, the first batch is only available here for now.

[lefix]

Your sharings are really well done !!! Can you share your last prompt ? I'm trying to force the ai tool to have limits or follow certain image composing instruction but after a lot of attempts i still can't have good results .....:Z

[fmoney963]

I love these.

[BBB]

Originally Posted by fmoney963

I love these.

Originally Posted by lefix

Your sharings are really well done !!!

Thanks! Your responses help motivate me.

Originally Posted by lefix

Can you share your last prompt ? I'm trying to force the ai tool to have limits or follow certain image composing instruction but after a lot of attempts i still can't have good results .....:Z

So I talked a bit about my prompt design in an earlier project. I'm most familiar at this point with Stable Diffusion and Flux. Flux understands more direction like "the woman on the left is short," while Stable Diffusion just didn't. My prompts for this set, as always, took a lot of trial and error. Here is a large part of the prompt I used for some of these (one reason there are so many attempts is because I'm constantly tweaking the prompt).

three part sequence of an attractive model's age progression transformation.
Version 1 (LEFT), the smallest, is fully clothed in cute intact pajamas. She has a petite, slender frame, 140 cm stature, immature face, weak shapeless chubby arms and legs, and tiny perky breast buds. She's focused inward.
Version 2 (CENTER) has a lanky, slightly taller body, 160 cm stature, and a somewhat developed figure with medium-sized bust, long slender arms and legs, slight curves, and sparse trimmed pubic hair. She looks down, regarding her own body, looking with surprise, intrigue, greed, and hopeful anticipation. Her body is shooting up, expanding with kinetic energy. Those same pajamas are too short and splitting at the seams, fluttering apart from her jiggling breasts and hips, leaving her exposed.
Version 3 (RIGHT), at full height of 180 cm stature, is shapely and filled out, with the matured face and body of an adult woman: large, full sized breasts and nipples, wider hips, fleshier toned arms and legs, a very full bush of abundant pubic hair, and armpit hair. She looks impressed, bold, proud, and triumphant. She has a pleased grin of self-satisfaction. She is mostly nude, with those same pajamas in tatters at the feet of Version 3 (RIGHT).
Their action poses highlight the contrasting size and shape of their bodies while preserving their shared genetics, identical outfits, consistent scale, and feminine appeal. The headline reads "Title Text" in bold letters. They're in the kitchen or dining room at breakfast time.
It's an unsigned 4k wallpaper-quality high resolution color illustration with intricate textures and details.
Overall, it's an absolute world-class masterpiece illustrated by a professional artist. It's an aesthetically pleasing illustration with impeccable attention to detail and impressive composition and color balance.
negative prompt:
Version 1 (LEFT) is topless. Version 1 (LEFT) is bottomless. They wear different-colored outfits. They're looking at each other making eye contact. Version 3 (RIGHT) is pear-shaped or obese with lipodema. The page includes text, labels, links, a signature, and other writing. The linework is busy or unfinished.

The first words have the most influence, which is why the part in red defining it as a 3-part sequence is first. I've noticed other people use "triptych", which is an art term to precisely describe a 3-panel painting. The more you can use art terms with specific, single meanings like this, the better (even though I didn't here, ha) since they can't be confused for something else. Some words like "attractive" just have good positive vibes, so even if the AI focuses on making the art style attractive when I was asking to make the character attractive, it's not a bad thing. Like I said earlier, getting the clothing transition was one of the hardest parts, which is why I mention "Version 3 (RIGHT)" a second time when describing where the tatters are. I figure the longer the text between two words, the less strongly the AI will associate them. I think "shared genetics" was an important find to describe a transformation, since "different versions of the same woman" could mean different in myriad ways you don't want. Lastly, a large proportion of the prompt is devoted to describing the character's emotion or facial expression in each panel. Having her react is a big part of making it a sequence progressing through time instead of just version 1, version 2, and version 3 that happen to be on the same page.

[BBB]

The third batch of girls drink their milk.

https://www.imagebam.com/view/GAGF4T

I find myself spending more and more time fixing up the imperfections on these in Photoshop. This is the first batch where I've written speech (it was easier than erasing all the speech bubbles!), and I even had to create half a leg that was missing. There are many ways the AI can give you almost what you want except for this one thing, especially with as much as there is going on in this concept.

[lefix]

Thank you very much !!! You really do an excellent work and helped a lot to understand the prompt composing and understanding. The ''earlier project'' was also well done , but i tried different prompts composition same as for stable and flux but still the lines was missing something. Looking forward for new result ill' share and wishing best regards to your excellent effort ---->thumb up