Some random AI images

[Dr.Minimizer]

After some very helpful conversations with the amazing MollySmalls, while working on the comic adaptation of Maggie & Jane, I decided to try out GoogleFX, and it's extremely good at producing some incredibly real images (as you can tell by looking at Molly's work). I can only imagine where I can take this technology in the future for comic style storytelling! In the meantime here are a few random images for you to enjoy...







I was thinking about maybe doing Magician's Bargain at some point...













Of course it wouldn't be me without some superheroine images...



























She looks really upset for some reason...

[Deadplay]

amazing work as always!! I've been a long time fan of your work ever since the old GTS/SW boards and your original website...

[Some Wanderer]

Great work, as usual. I've been using GoogleFX too, and it's pretty good at making SW art. It's surprising. Actually I've made some pics inspired by Magician's Bargain. I'll share them at some point.

[mrBoffkin]

I'm also trying to tame neural networks, but so far it's not working out well.

[Dr.Minimizer]

It's difficult to make the woman appear tiny. If you say she's tiny or small or minuscule, you get a plastic looking fake character. The trick that works for me is to focus on a gigantic object first. The AI appears to set the scale based on that object. So here are three examples:

First off I'll use the prompt below:

SpoilerShow

85mm telephoto overhead view of an athletic 23-year-old Caucasian woman with short black hair and darkly tanned legs wearing a black sleeveless bodysuit with white trim, black knee-high lace-up boots, solid black elbow-length opera gloves, a black collar, and a red and white belt with a large silver star on the buckle. She is covered in a sheen of moisture, and is standing on a gigantic black laboratory desktop that fills the entire background of the scene. On the surface of the desk, in the background, is a gigantic pen that fills most of the background of the scene, and several gigantic paper clips surrounded by enormous drops of water. She is looking around at the desktop in confusion.

Here's one of the pictures it gives me:



Next I'll change only a few words, replacing "is a gigantic pen that fills most of the background of the scene" to "a gigantic water bottle"...

SpoilerShow

85mm telephoto overhead view of an athletic 23-year-old Caucasian woman with short black hair and darkly tanned legs wearing a black sleeveless bodysuit with white trim, black knee-high lace-up boots, solid black elbow-length opera gloves, a black collar, and a red and white belt with a large silver star on the buckle. She is covered in a sheen of moisture, and is standing on a gigantic black laboratory desktop that fills the entire background of the scene. On the surface of the desk, in the background, is a gigantic water bottle and several paper clips surrounded by enormous drops of water. She is looking around at the desktop in confusion.

Finally I'll change "a gigantic water bottle" to "three gigantic paper clips so big they almost fill the scene"...

SpoilerShow

85mm telephoto overhead view of an athletic 23-year-old Caucasian woman with short black hair and darkly tanned legs wearing a black sleeveless bodysuit with white trim, black knee-high lace-up boots, solid black elbow-length opera gloves, a black collar, and a red and white belt with a large silver star on the buckle. She is covered in a sheen of moisture, and is standing on a gigantic black laboratory desktop that fills the entire background of the scene. On the surface of the desk, in the background, are three gigantic paper clips so big they almost fill the scene, surrounded by enormous drops of water. She is looking around at the desktop in confusion.

These were all done in portrait mode. Notice I never once mention the size of the woman, even in passing. She is described using ordinary words that don't mention scale. As soon as you start doing that, you start getting fake looking images. The trick is to have the AI zoom in on the objects around her. Of course that means if you want to do something like have her running across a blank floor, it's not going to work. You can either add a giant object nearby and then crop it out or have an AI tool remove it, or use Photoshop to put her into such a scene. The AI is great but can't do everything quite yet.

[eddiegiantman]

How about Sasha's Weight Loss or Too Much of Anything?

Fire and Ice?

[Dr.Minimizer]

Fire and Ice has some real possibilities. The AI won't let me do anything with any kind of sexual overtones, though, so there won't be any fondling or anything like that. There's a certain scene in Magician's Bargain, for example, that would also be very difficult to do. Sasha's is a pretty weak story.

Before I go rehashing any other stories, there's going to be one using that girl in black I keep posting images of. Stay tuned. Here's another preview...

[MollySmalls]

Thank you for the kind words! Blown away by what you've created so far, looking forward to more of your works. Your style is so distinctive and it's amazing to see it come to life with AI.

[Sumguy14]

thanks for including some of your prompts. I'm messing around with it as well and that really helped! My initial efforts have been a mixed bag but some of them turned out kind of good.

[Hand-Holder]

They may be random but they are spectacularrrrrrrrrrrr

[Dr.Minimizer]

I still can't get the AI to let me make a giant hand actually grab and hold onto a tiny woman. I can get her to sit in the palm in various poses, but the AI doesn't listen when I tell it where to place the giant fingers. Bummer.

I did at least learn to prompt by putting the initial focus on the folds and ridges of the palm, which helped make the girl look smaller without totally turning plastic. Enjoy...

[Some Wanderer]

Yeah I've tried making some handheld pics too but with no luck. I can't figure out a way to implement the concept of a handheld into a prompt.

[dubiouskey]

Good shit!

[Dr.Minimizer]

It's all about how you describe the background. The AI focuses on the character, then kind of wraps the rest of the scene around that character. So if you write this...

85mm telephoto DSLR photo front view of an attractive athletic young Caucasian woman wearing a black bikini. Seen from above, she is standing, looking up, with arms held out and pleading. She is soaking wet and is standing on a gigantic man's palm.

...you get this...



Change it to

85mm telephoto DSLR photo front view of an attractive athletic young Caucasian woman wearing a black bikini. Seen from above, she is standing, looking up, with arms held out and pleading. She is soaking wet and is standing on a gigantic man's palm that fills the entire background of the scene.

...and the hand doubles in size...



Add another object...

85mm telephoto DSLR photo front view of an attractive athletic young Caucasian woman wearing a black bikini. Seen from above, she is standing, looking up, with arms held out and pleading. She is soaking wet and is standing on a gigantic man's palm next to one of the fingers. There is a wedding ring on the finger that fills most of the background of the scene.

...and the background gets even bigger, even if the AI doesn't quite put the ring where it should be.



Then just for fun you can go crazy...

85mm telephoto DSLR photo front view of an attractive athletic young Caucasian woman wearing a black bikini. Seen from above, she is standing, looking up, with arms held out and pleading. She is soaking wet. The background consists of a single set of fingerprints on a man's fingertip that fill the entire background of the scene.

[potyzeff]

these look great!