Viridian's AI Experiments

Artificial Intelligence is here! Really! Anything created with AI assistance, including stories, should be posted here.
Viridian
Posts: 1581
Joined: Wed Apr 15, 2009 10:03 am

Re: Viridian's AI Experiments

Postby Viridian » Tue Jan 31, 2023 1:37 pm

An AI remake of my "Emergency Dispatch" drawing.
You do not have the required permissions to view the files attached to this post.
Viridian @ deviantART: http://viridianqs.deviantart.com/

User avatar
sixgunzloaded
Posts: 916
Joined: Tue May 05, 2015 8:16 pm

Re: Viridian's AI Experiments

Postby sixgunzloaded » Tue Jan 31, 2023 8:52 pm

Viridian wrote:Here's a Susan-inspired piece.


Oh wow! I just found this today! I love this! haha! :D
Actually, I love them all but, as you can probably tell, I'm particularly partial to "Susan". :lol:
One of the things I'm really liking about this AI technique is the sort of softness in all the images. If this were Photoshop, I'd call it "Diffuse Glow". Not sure if that's the right term here but that's what it reminds me of.
"Cellphone" is really nice too. I like the dispatcher's secret reaction! :lol:
Well done, and thank you! Especially for "Susan". Please feel free to use her likeness as much as you like! :lol: :P :D
How long did Tarzan watch before deciding to save Jill..?

SPACEMARINE394
Posts: 93
Joined: Sat Dec 12, 2020 1:09 am
Location: WI
Contact:

Re: Viridian's AI Experiments

Postby SPACEMARINE394 » Thu Feb 02, 2023 11:58 pm

Very nice. Does anyone have a link to the David's Day story you mentioned?
FOR THE EMPEROR

Viridian
Posts: 1581
Joined: Wed Apr 15, 2009 10:03 am

Re: Viridian's AI Experiments

Postby Viridian » Fri Feb 03, 2023 10:20 am

sixgunzloaded wrote:
Viridian wrote:Here's a Susan-inspired piece.


Oh wow! I just found this today! I love this! haha! :D
Actually, I love them all but, as you can probably tell, I'm particularly partial to "Susan". :lol:
One of the things I'm really liking about this AI technique is the sort of softness in all the images. If this were Photoshop, I'd call it "Diffuse Glow". Not sure if that's the right term here but that's what it reminds me of.
"Cellphone" is really nice too. I like the dispatcher's secret reaction! :lol:
Well done, and thank you! Especially for "Susan". Please feel free to use her likeness as much as you like! :lol: :P :D

A few more Susan-inspired pieces.
You do not have the required permissions to view the files attached to this post.
Viridian @ deviantART: http://viridianqs.deviantart.com/

User avatar
sixgunzloaded
Posts: 916
Joined: Tue May 05, 2015 8:16 pm

Re: Viridian's AI Experiments

Postby sixgunzloaded » Fri Feb 03, 2023 9:25 pm

Viridian wrote:A few more Susan-inspired pieces.


Geez, these look amazing. :) Particularly, "Final Trap". It looks like you've pretty much got the expression aspect of this figured out. Especially the eyes. They suggest anger at the situation she finds herself in, as well as fear.
The texture and moisture on her skin (in all 3 pics) looks really nice as well. The hair has a nice wet look to it, like maybe she fell into the cistern before she fell in the qs. lol Good touch. The qs has a good clumpy, piled-up appearance, and I think it's some of the best yet.

"Horrible Mud" looks great for the most part too. It has that dark woods setting that I was trying for in the original and also, the way the mud blends into the background. The only thing I think looks a little off are the eyes in this one. They look sort of... missing. :lol: I'm gonna assume this is just how the program interpreted it. ;)

"Nightmare In Tar" looks like it could have come right out of my own computer - only I think yours actually looks better! The tar is perfect! It has that slimy black sheen that I haven't been able to get quite right yet, myself. Her dress is also nicely re-created and I love the definition of all her body parts.

Really nicely done with these, thank you!

You've got me curious now about this software. I've always liked seeing updated versions or new or different takes on existing materials. Seems like this program could theoretically generate endless variants on stuff, depending on the user's input. That could be a fun experiment all on its own.
How long did Tarzan watch before deciding to save Jill..?

User avatar
Theo
Posts: 684
Joined: Sat Feb 13, 2021 4:12 am
Location: Sinking Fantasyland

Re: Viridian's AI Experiments

Postby Theo » Fri Feb 03, 2023 10:30 pm

At first I wasn't too sure about AI quicksand art. But they're starting to turn out pretty dam good! :P
Finally broke down and went to see a psychologist. She told me I have an overactive imagination—and it really excited her. We're going sinking next weekend. Theo's AI Quicksandbox New stuff every weekend (unless life gets in the way)

Viridian
Posts: 1581
Joined: Wed Apr 15, 2009 10:03 am

Re: Viridian's AI Experiments

Postby Viridian » Sat Feb 04, 2023 11:51 am

sixgunzloaded wrote:"Horrible Mud" looks great for the most part too. It has that dark woods setting that I was trying for in the original and also, the way the mud blends into the background. The only thing I think looks a little off are the eyes in this one. They look sort of... missing. :lol: I'm gonna assume this is just how the program interpreted it. ;)

This was one of the challenges in working with the originals. While the output was high-res, the model details are somewhat pixellated, especially around the cheeks and eyes. I wasn't sure myself what the original scene was portraying, whether it was just the shadows and/or mud in her face, or whether she was being possessed by the nightmare, hence I let the AI leave it as a shadow. However, adding eyes is an easy fix.
horriblemud1.png

You've got me curious now about this software. I've always liked seeing updated versions or new or different takes on existing materials. Seems like this program could theoretically generate endless variants on stuff, depending on the user's input. That could be a fun experiment all on its own.

That's one of the great advantages of AI tools. Here's the same scene with a comic twist and a manga twist.
horriblemud2.png
horriblemud3.png


Theo wrote:At first I wasn't too sure about AI quicksand art. But they're starting to turn out pretty dam good! :P

I've said earlier that it's easy to knock off low quality QS generated art if you don't know what you're doing. But once you've learned what to input, you can get some really, really good pics that can perfectly match what you envision.
moments_qsoasis.png
You do not have the required permissions to view the files attached to this post.
Viridian @ deviantART: http://viridianqs.deviantart.com/

User avatar
Theo
Posts: 684
Joined: Sat Feb 13, 2021 4:12 am
Location: Sinking Fantasyland

Re: Viridian's AI Experiments

Postby Theo » Sat Feb 04, 2023 9:30 pm

Cool! I think I prefer her expression and way she's looking over her shoulder in the original, but it's not too far off. I still don't entirely understand how it works (mainly because I haven't used it much). Do you upload or link a picture as a reference, or is it all just based on your textual input?
Finally broke down and went to see a psychologist. She told me I have an overactive imagination—and it really excited her. We're going sinking next weekend. Theo's AI Quicksandbox New stuff every weekend (unless life gets in the way)

Viridian
Posts: 1581
Joined: Wed Apr 15, 2009 10:03 am

Re: Viridian's AI Experiments

Postby Viridian » Sun Feb 05, 2023 1:10 am

Theo wrote:Cool! I think I prefer her expression and way she's looking over her shoulder in the original, but it's not too far off. I still don't entirely understand how it works (mainly because I haven't used it much). Do you upload or link a picture as a reference, or is it all just based on your textual input?

Of course, you can have it any way you want. I just picked a scene and added the text afterwards. I'm not the original creator, so I don't have the context and intent at the time it was made, and that's what makes AI art still very much human-driven: you're the one who adds the meaning to the illustration. The objective wasn't to re-create the scene 1:1, but to show how AI can interpret the scene.

AI tools can use both image-to-prompt and text-to-prompt. In simple terms, imagine that you're asking someone to illustrate a scene for you. You have to explain to the illustrator what you want and what specific things need to look like (e.g. girl in quicksand, long red hair, wearing blue tank top, bit of cleavage, etc.). The thing, of course, is that all these elements are still vague and up for interpretation or misinterpretation, so part of the process is to get it just right. You would tell your illustrator to change the shade of blue or the length of hair. This is basically what you're doing with AI. Equally, if you give your illustrator a picture reference, it is a far clearer idea of what you want and you get more accurate results, and you can make adjustments.

Here's an example of how I output an image. I could either make the AI generate a pose that suits my product, or I can use an image reference. In this case, I'm using this photo of model Becky Hudson, because it'd be nice to see this specific outfit and pose in quicksand.
Step1.png

Similar to how photomanips use Photoshop to chop out backgrounds and paste the model into quicksand, I need to shape the piece to look like how I want it.I crop the image and do a basic paint of the scene's elements - the sand and sky. For the sake of this demo, I'm also going to change the hair colour to grey, so I'm painting over that. Notably, I also add in ripple lines, because the AI won't recognise what I have in my mind when I think "quicksand", and I want this one to have thick oozing ripples. So now it looks like a grade-schooler's first drawing.
Step2.png

I then specify the prompts and tags. These are guides (more so than instructions) on what the AI should be including. The more specific you are with the prompts, the more the AI understands what it's looking at. So things like "quicksand" make more sense with the rippling pattern I made. I specify "grey hair" (otherwise it thinks it's a veil), and also add in "beach" and "sky" tags. I want to keep the eyes closed, but I also want the subject to be more scared or worried, so I add "smile" to negative prompts to remove the ambiguous expression. And I'll add some nippleage because boobs. I'll generate a few and pick the one that I like most.
Step3.png

This alone is looking really, really good. The basic subject matter is all there, and the quicksand turned out REALLY good. This is where a lot of creators might jump the gun and upload this as finished, but I'm not 100% satisfied that the styles blend together and some details could be further refined. However, this image provides a much closer vision to the AI than what I started with, so I can either paint more details over it, or add in more prompts. In this case, I wanted to add "sun" and "clouds". There was no sun reference in the image, so that didn't really show, but the clouds were added nicely, and with the final generated image, the AI makes more sense of the scene and makes the hair consistent, plus adds elements like water/sweat to match the interaction with what it now understands as a hot day on the beach, in quicksand.
Step4.png
You do not have the required permissions to view the files attached to this post.
Viridian @ deviantART: http://viridianqs.deviantart.com/

Viridian
Posts: 1581
Joined: Wed Apr 15, 2009 10:03 am

Re: Viridian's AI Experiments

Postby Viridian » Sun Feb 05, 2023 4:10 am

This time, no image reference. This is straight text-to-prompt. I use the exact same prompt as the previous set to see what the AI can spit out (grey hair, white bikini, large breasts, nipples, beach, sinking in quicksand, etc.). This is the first result.
Step1.png

Not bad, but too watery. So I paint over it and add the ripples and textures I want.
Step2.png

Then it outputs this graphic with the same prompts:
Step3.png
You do not have the required permissions to view the files attached to this post.
Viridian @ deviantART: http://viridianqs.deviantart.com/


Return to “AI Creations”

Who is online

Users browsing this forum: No registered users and 1 guest