AI generated zako (1 Viewer)

aesopn

Master of this Domain
Joined
Jul 21, 2012
View attachment 147062
My first experiment with AI greem; a Final Vendetta Karen.

Thanks for the inspiration!
I think it would be easier to generate a zako art without throwing in the original sprites.

Here is my attempt:
00366-277542985.png

Prompt: a buffed asian girl with blue sports bra and blue short shorts with boxing gloves, purple hair, medium hair, full body portrait, cleavage, firm lighting, open chest, midriff

Then mixed with some in-paint jobs to get the other color right (e.g. pink bit in hair, red boxing gloves...)

A good tip is to not include more than two color at once or the AI will start getting confused.
Try to just start with one color and polish it with in-paint afterwards.
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
Some more beat'em up zako I've try to recreate in real-life form

Jin - Eojjeonji Joheun Il-i Saenggil Geot Gateun Jeonyeok​

1687231316699.png
00279-423698744.png

Female Soldier - Dark Judgement​

1687231564672.png
00154-943519838.png

Finally, I have re-visited one of my favorite zako!


Girl - Jjang the young boss
1687231754371.png
00473-744346244.png

This pair of girls ambush the hero from behind at the start of stage 2, what a brave duo!
 
Last edited:

Tsukachan

Master of this Domain
Joined
Oct 1, 2017
Thanks for the inspiration!
I think it would be easier to generate a zako art without throwing in the original sprites.

Here is my attempt:
View attachment 147069
Prompt: a buffed asian girl with blue sports bra and blue short shorts with boxing gloves, purple hair, medium hair, full body portrait, cleavage, firm lighting, open chest, midriff

Then mixed with some in-paint jobs to get the other color right (e.g. pink bit in hair, red boxing gloves...)

A good tip is to not include more than two color at once or the AI will start getting confused.
Try to just start with one color and polish it with in-paint afterwards.

She looks very good!!
 

Tsukachan

Master of this Domain
Joined
Oct 1, 2017
Thanks for the inspiration!
I think it would be easier to generate a zako art without throwing in the original sprites.

Here is my attempt:
View attachment 147069
Prompt: a buffed asian girl with blue sports bra and blue short shorts with boxing gloves, purple hair, medium hair, full body portrait, cleavage, firm lighting, open chest, midriff

Then mixed with some in-paint jobs to get the other color right (e.g. pink bit in hair, red boxing gloves...)

A good tip is to not include more than two color at once or the AI will start getting confused.
Try to just start with one color and polish it with in-paint afterwards.

BTW, A aesopn
What model did you used?
I tried your prompt on Seaart.ai, but never got exact one
 

Watt Khan

Club Regular
Joined
Jul 6, 2021
chestnut-hair-buxom-chest-pointy-nips-japanese-smirk-feathered-punk-light-...-chestnut-hair-gr...jpg
IMG_20230630_003345.png


Another bit of AI-generated shenanigan. You think you can tango with Onihime/Mona and her twin sister Yasha/Lisa? I think it's a shame how criminaly underrated these girls are. I wish that the SOR fandom could give them a bit more love. Their absence in the latest SOR game (outside of concept art) is just disappointing.

I won't bother dropping the prompts for the image. The image you see now was only obtained after a fuckton of image regenerations plus prompt edits.
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
View attachment 147496View attachment 147498

Another bit of AI-generated shenanigan. You think you can tango with Onihime/Mona and her twin sister Yasha/Lisa? I think it's a shame how criminaly underrated these girls are. I wish that the SOR fandom could give them a bit more love. Their absence in the latest SOR game (outside of concept art) is just disappointing.

I won't bother dropping the prompts for the image. The image you see now was only obtained after a fuckton of image regenerations plus prompt edits.
I tried to generate Mona & Lisa in their SOR1 design.

It's surprisingly hard ngl, hope you like it!

And likewise, I've done around 1000 generations on this one in total,
each time tweak them a little (e.g. boots too long, wrong leather jacket, a t-shirt generated inside etc.)

Those headbands were surprisingly hard to generate, as I don't think the model recognized how a sport headband looks like very well.

In the end getting the right image was surprisingly rewarding though!

00188-122513539.png
1688258857993.png
1688258876587.png
 

Ultimatelegendsasa

Avid Affiliate
Joined
Jul 26, 2021
I tried to generate Mona & Lisa in their SOR1 design.

It's surprisingly hard ngl, hope you like it!

And likewise, I've done around 1000 generations on this one in total,
each time tweak them a little (e.g. boots too long, wrong leather jacket, a t-shirt generated inside etc.)

Those headbands were surprisingly hard to generate, as I don't think the model recognized how a sport headband looks like very well.

In the end getting the right image was surprisingly rewarding though!

View attachment 147655View attachment 147656View attachment 147657
what program do you even use for this. Crazy dude…. I want to try out what you pulled off too.
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
what program do you even use for this. Crazy dude…. I want to try out what you pulled off too.
I only used Stable Diffusion honestly, have it running on my computer (you do need a decent graphic card though, VRAM>5GB).

Used a realistic model (e.g. you can see some sample and choose the one you like here

I used a model called ICBINP (I can't Believe it's not photography).
But there is new and better model came out everyday, definitely have a look on a few examples and give them a try.
tip: You generally want a model that could generate wide range of various real live photos, man & woman

That's a starting point, then there's a few technology if you go further along.
After you generate a decent image that is 60% of your liking from Text to Image,
Try img2img to refine the whole image or in-paint to refine a single part of the image.

It is especially useful with like an image editor, just paint over a part you don't like and use img2img/in-paint to regenerate the whole thing.

Then there's a few technology with extra plugins if you wanna go EVEN further.
Like ControlNet, if you want to generate all the image with a specific pose, or you want to change the color but not the shape of an image.
And the most useful one I've found recently, ADetailer, which essentially repaint the whole face as they get distorted very easily especially when you generate a full body image.


I can go on and on, it's mainly depending on how determined you are to generate your dream image.
Like myself could spend hours polishing just one perfect image,
while some others just like to mass generate 1000 crappy image that made themselves happy.

Just node that a good image is not as simple as finding a magic model and BOOMM, reward!
 

Ultimatelegendsasa

Avid Affiliate
Joined
Jul 26, 2021
I only used Stable Diffusion honestly, have it running on my computer (you do need a decent graphic card though, VRAM>5GB).

Used a realistic model (e.g. you can see some sample and choose the one you like here

I used a model called ICBINP (I can't Believe it's not photography).
But there is new and better model came out everyday, definitely have a look on a few examples and give them a try.
tip: You generally want a model that could generate wide range of various real live photos, man & woman

That's a starting point, then there's a few technology if you go further along.
After you generate a decent image that is 60% of your liking from Text to Image,
Try img2img to refine the whole image or in-paint to refine a single part of the image.

It is especially useful with like an image editor, just paint over a part you don't like and use img2img/in-paint to regenerate the whole thing.

Then there's a few technology with extra plugins if you wanna go EVEN further.
Like ControlNet, if you want to generate all the image with a specific pose, or you want to change the color but not the shape of an image.
And the most useful one I've found recently, ADetailer, which essentially repaint the whole face as they get distorted very easily especially when you generate a full body image.


I can go on and on, it's mainly depending on how determined you are to generate your dream image.
Like myself could spend hours polishing just one perfect image,
while some others just like to mass generate 1000 crappy image that made themselves happy.

Just node that a good image is not as simple as finding a magic model and BOOMM, reward!
Can you try making the Kabuki Twins from Batman?? they seem pretty doable, maybe the mask will be difficult. I’ll try doing it as well!! :D
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
Can you try making the Kabuki Twins from Batman?? they seem pretty doable, maybe the mask will be difficult. I’ll try doing it as well!! :D
There is a good reason why I avoid generating weapon as much as possible.

All AI art models so far can't generate a decent weapon, all character can't even hold a knife properly, and most items are in weird shape too.
For this character, definitely really hard consider the weapon precision needed for this character.
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
Capture3.PNG


Ace Attorney Inspired scene: Case of the Trainyard explosion

Capture1.1.PNG


Capture2.PNG


Imagine a scenario where Doll luckily enough to escape the wrath of the angry rangers and the case was brought to court :D

Thought it was a fun little crossover plus testing out a new model.

I had fun making it, hope you guys like it too.
 

aesopn

Master of this Domain
Joined
Jul 21, 2012
Battle Circuit: Bike Gal
1691636913542.png

00345-109464057.png
00497-2813924049.png


Police seems to caught a young girl speeding!

"I recommend you to stay out of this" She said. "Otherwise you gonna be in big trouble."

The hardest part of the battle circuit gals is to keep their jacket just enough to cover their breast, and not generate a bra on top of it.

Me myself really like these sort of thing, the real sexy is right about that 1% from fully exposing.


I like the anime model because it made good expressions. But I do see why people enjoy the realistic model more tho :)
 

Users who are viewing this thread

Top


Are you 18 or older?

This website requires you to be 18 years of age or older. Please verify your age to view the content, or click Exit to leave.