Keeper live action version

[Dr.Minimizer]

I've always wanted to do a live action version of Keeper. (The comic can be found here if you haven't seen it or don't remember.) In fact I even once almost paid a model to pose for photos to make it happen, but that would've been over a thousand dollars, and didn't include the other characters in the story, so I couldn't afford it and gave up. Now, though, AI says hi...and here's what it came up with for the cover (with a little Photoshop on the assist)...



BTW here's what the AI came up with for Alison's look...

[Some Wanderer]

Looks good.

[Firewall]

I look forward to this. Funny thing is that my story "Shrink to Fit" was inspired by your "Choose your shrinking" story from back in the day.

[Dr.Minimizer]

That's the best thing about having a community...ideas bounce off each other, and lead to more creativity and inspiration.

[DocRick]

Looks like the same tiny in the first picture.

[Dr.Minimizer]

I couldn't get the AI to depict Alison with the coat buttoned shut, so it's come open in the fourth panel, but Troy is so fixated on the portal opening that he hasn't noticed yet (but he will).

Original hand-drawn page 1: https://www.deviantart.com/the-minimize ... -727007373

[theangelofdeath1986]

Awesome

[Dr.Minimizer]

So far the AI is giving me fits with the buckled boots and getting the belt to go around her hips and not her waist, but as yet I have not had to resort to any Photoshopping. That will change in the next page I'm sure. Also, I'm using the exact same dialogue so far, and for the moment it's a shot for shot rermake, but I can't guarantee that will remain the case throughout.

Original hand-drawn page 2: https://www.deviantart.com/the-minimize ... -727162778

[Dr.Minimizer]

Of course I knew I could never get that last panel exactly right, but I got as close as possible. I actually described the six ladies in the image, and AI always added more, no matter what I did. Plus, if you use certain words, like bikini or leotard, it often refuses to make the image. Add several of them and you're not going to get anything. I think I tried at least a couple hundred times and got maybe 10 images to even render at all, and the ones that did usually had some kind of flaw, like a guy in the image, or women with their heads on backward, and stuff like that. Honestly I think I was lucky to get anything.

Original hand-drawn page 3: https://www.deviantart.com/the-minimize ... -727200730

[Dr.Minimizer]

Couldn't quite figure out how to do the thought sequence with the AI art generator, so I came up with this for the last panel. At least it put the ladies in the right places around the table.

Original hand-drawn page 4: https://www.deviantart.com/the-minimize ... -727494772

[Dr.Minimizer]

The AI refuses to draw any kind of giant power cords that are thicker than a couple of inches wide, so I settled for white plastic curving pipes. Whatever works!

Original hand-drawn page 5:https://www.deviantart.com/the-minimize ... -727602370

[Dr.Minimizer]

So the AI absolutely refused to portray the handset anywhere other than on the phone, and if I tried to get it to be on the desk by itself, it was either face down, or else if it was face up it was covered with phone buttons (like the AI merged the faceplate of a phone with the handset). So I gave up and had to redo this scene slightly with Alison using the speaker mode to make the call. Oh well, it gave me more opportunities to show her struggling to do something that would ordinarily be trivial, plus you all get to enjoy an extra page. Win-win!

Original hand-drawn page 6:https://www.deviantart.com/the-minimize ... -727831817

[Little Sally]

(images)

So the AI absolutely refused to portray the handset anywhere other than on the phone, and if I tried to get it to be on the desk by itself, it was either face down, or else if it was face up it was covered with phone buttons (like the AI merged the faceplate of a phone with the handset). So I gave up and had to redo this scene slightly with Alison using the speaker mode to make the call. Oh well, it gave me more opportunities to show her struggling to do something that would ordinarily be trivial, plus you all get to enjoy an extra page. Win-win!

Original hand-drawn page 6:https://www.deviantart.com/the-minimize ... -727831817

Can't work out why you're not getting more feedback on this, it's extremely well done and entertaining (even though our size preferences differ).

Not being a user of AI, I can only imagine the hard work you've put in creating this, and it's great you've shared the tricky things you've had to do.
But what you've done is working, so well done!

[eddiegiantman]

How hard is it to maintain the title characters look?

I would ask to see if it can redo Diminishing Opposition but let's not go all in.

[Dr.Minimizer]

Well, tell you what, you can try it yourself. This is all done with Google ImageFX, which is free to use up to a certain number of generations per day (which I hit frequently, lol). Here's the basic prompt and then I'll post the first four images that come out of the AI generator so you can get an idea of the differences that can pop up.

Prompt:

SpoilerShow

85mm telephoto DSLR photo high angle view of a smiling woman standing with one hand on her hip and the other waving as she looks up toward the top of the scene. She is an 18-year-old slim buxom Caucasian woman with short blonde hair in a pageboy style and olive skin, no makeup or piercings, wearing a purple glittery sleeveless scoop neck crop top with black trim, a purple collar, a wide purple hip-hugging belt slung low around her hips, black bikini bottoms, black fingerless gloves, fishnet pantyhose that cover all of her legs, and black buckled knee-high boots.

Images:



I used some of MollySmall's suggestions on how to design prompts to get some highly realistic imagery. In this particular image I didn't describe any background info so it just put her wherever it wanted, in this case in a street for some reason, but if you wanted to add background stuff you can put it after the description.

Generally the more detailed you get, the more the AI will try to meet your requests, and if you don't say something then it will do whatever it wants. So if I didn't say slim, she would sometimes come in varying body shapes. I like to say "buxom" but most of the time you can't tell the difference. If you try to get specific like say DD-cup or huge breasts, most of the time the AI just fails to make the image entirely. Google ImageFX won't do nudity or anything it considers anywhere near pornographic, so even something innocuous like "low-cut" will fail, and depending on the pose, sometimes it won't even do fishnets. Notice I had to say "fishnets that cover all her legs" because if I don't they will be more like socks half the time, or be incomplete, or be super sparse that doesn't give me the result I want.

Notice in the four images all the things that are different despite me stating otherwise: the hairstyle, the belt, the boots in the fourth panel (most of the time it's 50-50 whether I get thigh boots when I say knee-highs), etc. So quite often it just does its own thing which can be frustrating. But also notice the differences where I didn't get specific: the cross hanging from her collar, the lacy gloves, the big hoop earrings, the angle on the belt, the flaws in the fishnets, and so on. If those things are a problem, alter the prompt to change them, but just be aware if you get too much information there, the AI will sometimes refuse to make the image at all, especially with highly complex backgrounds. When I do multiple characters, I often have to just be more vague and live with whatever it gives me.

Usually when I go for an image it will have 3 or 4 things that aren't working properly and I refine them with more adjectives or commands, usually related to the pose or shot angle, but even so I often find myself running 10 or more generations just hoping to get that one where everything lands perfectly. It can be frustrating and you need a LOT of patience. I can see why MollySmalls had to spend so much time and effort on this, and it makes me appreciate Maggie & Jane all the more!

[Dr.Minimizer]

I couldn't get the AI to generate a rear view of Katie in the third panel, because for some reason it doesn't like drawing backsides very much, often flaking out when I try it, even with only a single character in the scene. So you'll have to wait for the butt shot of Katie until later in the comic. Sorry!

Original hand-drawn page 7: https://www.deviantart.com/the-minimize ... -727960769

[Dr.Minimizer]

This is where the fun stuff starts. Unfortunately the AI is going to be a huge pain with the handheld scenes. For some reason I can get it to do a woman held in a man's hand, but not in a woman's hand. So it's all Photoshop from here, most likely. Oh well, live and learn.

Original hand-drawn page 8: https://www.deviantart.com/the-minimize ... -728198126

[scidram]

This is so much fun! Loved the original comic and its concept, and seeing it like this is great! And wow, now that you've explained some of your process, I'm amazed at how much work is going into this! Thanks for sharing!

[theangelofdeath1986]

Loving this. AI can be used for good it seems.

[Dr.Minimizer]

Okay so I made two versions of the next page, one zoomed in like you see in the original and the other zoomed out to give a better idea of the scale difference. Which one do you all prefer? Feel free to chime in with comments or a private message.





So which do you like, the zoomed in or zoomed out version?

Original hand-drawn page 9: https://www.deviantart.com/the-minimize ... -728296031

[Dr.Minimizer]

Wow, nobody has an opinion? Okay, whatever, moving on then...



There are some things the AI won't even dare to draw, and that last panel is one of them, but that's what Photoshop is for. (Actually I used Photoshop to make 4 of these 5 panels, but all the different pieces were made with the same AI.)

Original hand-drawn page 10: https://www.deviantart.com/the-minimize ... -728297766

[eddiegiantman]

Wow, nobody has an opinion? Okay, whatever, moving on then...



There are some things the AI won't even dare to draw, and that last panel is one of them, but that's what Photoshop is for. (Actually I used Photoshop to make 4 of these 5 panels, but all the different pieces were made with the same AI.)

Original hand-drawn page 10: https://www.deviantart.com/the-minimize ... -728297766

I do like how you were replicating the angles of the original.

[Dr.Minimizer]

It's not always possible, because there are only so many camera angles you can command it to use, and sometimes it outright refuses to draw anything if it doesn't want to, without ever explaining why. For example in panel 3, it absolutely refused every attempt to make that rear side pose, crashing every time with zero renders, until I told it to have her looking up at a bird flying in the sky in the back right corner of the scene. Then it gave me that view of her, after about 15 tries. 2 hours of attempts just for that one pose. *sigh*

Anyway here's another page...



Original hand-drawn page 11: https://www.deviantart.com/the-minimize ... -728622406

[eddiegiantman]

PG. 12 will probably be a tough one to redo.

[Dr.Minimizer]

This one was the hardest. There were six panels, and nearly all of them gave me fits and they all required Photoshop to add Alison into the scene. Honestly #4 was probably the easiest. I probably ran 75+ renders on panel 5 before I got both the ladies right at the same time. The AI always changed one detail or another. (I love Katie's outfit, and maybe I'll shrink her too if I ever do a sequel.)

Original hand-drawn page 12: https://www.deviantart.com/the-minimize ... -728622433