🤯 Any other AI art addicts here? 🤯

[J.L.O]

Well, this is more demented than pop art but yeah.

[Kamiccola]

I think I’m going to cave and buy myself more credits because 5 a day isn’t nearly enough. I’ve created so many gorgeous Amphoras, I can’t contain how happy it makes me to see the AI respond to prompts this well.

My latest fave is this one. A very simple portrait that doesn’t represent her fantasy race at all but she’s gorgeous and perfect.

[NotARussianBot]

Accuracy what editing is for

[Kamiccola]

I was trying to create Lamassu, who’s a winged big cat, and found that it’s difficult to create a winged cat. But this is my favorite so far. I used it as my phone wallpaper.

I found that pop culture gets in the way of good creation.
For example, when trying a prompt with “black panther,” it kept giving me superhero costumes. I thought, well, puma is another word for a panther, I tried that and the AI was thoroughly confused, didn’t know what to paint. I guess there weren’t enough puma (cat) images used to train it.

I finally got better results with “Black lioness” prompts. Still hard to put wings on my cat but it’s starting to look cool.

[Kamiccola]

Today I made a sad observation.

I was trying to create an image of a man tied up laying on the floor to reflect one of the scenes in my book but AI struggled imagining what that is, kept producing very strange anatomy, so I googled it to maybe feed AI an image to give it an idea of what I want.

And Google image result is what makes me sad. I couldn’t find any images of a tied up man. I found a couple of pics of a man tied to a chair but nothing like what I needed.

Instead I found a shit ton of stock images of tied up boys.
Why TF does anyone need that many stock images of preadolescent boys tied up and helpless on the ground or cowering in the corners?

The only explanation I can think of makes me furious and I just give up.

I don’t want this prompt anymore.
Let me make a cute cat instead.

[NotARussianBot]

Doom has some background sprites that might be useful-there is a rumor that it was based on pictures of Mussolini and his followers’ ultimate fate.

It’s disturbing but they are adults.

[J.L.O]

Ok, so i'm going to show you what I'm doing in Dream to deal with "Demented AI issues".

So, the prompt is “male butler running away from Pearl Harbor”. The responses are all over the place, but the better the render, the more likely the pose is going to look like they have no action:

This is one of the renders that never gives me proper details at all, but at least looks like action is going on:

So I use this crazy as the input image and usually have it set on normal. If it’s vague shape is what I want to see at the end, I bump it higher.

This one is the Surreal render. It usually throws up alien spaces, and smooths out things unless I put it under prompt like this, and this is the only one so far that looks like we are looking down on parked boats.

This one is HD, and usually gives me the most accurate details.

This one is Retro Futurism. Screwed up the face, but is better than the original pic.

And Street Art, where some screw-ups looks like artistic license.

The whole point is you take the demented sneeze from an over-achieving AI model and feed it back to a smoothing AI model, with the same word prompts, and see how it handles it all. I’ll get stuff all over the place, but eventually, I get one that’s like “Why didn’t you give me this in the first place?!”

[J.L.O]

So, when I’m killing time, I’m setting up dummy covers for stuff I don’t need…yes, through the messy-then-smooth plus editing.

[Kamiccola]

I’m going to get a few nice images into Shutterstock.
I figured I might as well have use of the images that I won’t use otherwise. And someone out there could be looking for something just like what I’ve created.

I started working on it last night. I used one of the rejects in the quest for Amphora image. I upscaled it all the way in NightCafe to get a higher resolution and then hopped on over to Krita to make some corrections. I gave her pretty eyes and smoothed out a few spots (mostly with a clone tool). But I can’t figure out how to fix the lip shape in Krita.

So I think I’m going to have to use a different app to fix the lip, maybe GIMP?

And that will be the last step. I’ll submit it and if it’s approved then I’ll call myself a professional digital artist.

[J.L.O]

Warp is your friend, if it’s subtle. I’ve just reconstructed lips altogether. For example, that umbrella chic, her lips are 4 dots and a smudge.

[Kamiccola]

Yeah. I navigated to Krita because it’s an art tool and I’m working on art, right? But considering that I mostly used only clone and eraser tools and just need an ability to do layers and filters and blending, then an editing tool might be a better option.

For reference, using Glow blend on the eye layer worked beautifully. It’s hard to tell that those eyes weren’t part of the image to start with but pretty eyes really upscale the image to another level. It takes some creative tweaking though to blend them in smoothly.

You know, it occurs to me that I would fail miserably if I tried to make a living from digital art because I spend hours on one image.
I read someone say that from having 2000 images on a stock site, they get about $200 a month in commission and that’s a pretty successful result.
At the speed I create these, it would take me a year to create this many images, and well, can’t live on $200 a year.

Luckily, I enjoy the process. My only hopes are that I could make enough commissions from stock to justify subscription to one (or more) of the AI programs so I could play with creation without limits. My primary use is still to create art for my own projects. The ones that would go to stock would be the ones that came out nice but not what I wanted.

[J.L.O]

Starting from edits, it takes maybe an hour to finalize something that was mostly finished. If it’s a simple thing to put together, only ever an hour…had some things come together in 20 minutes. But other tings that just don’t work take days.

[Kamiccola]

https://www.instagram.com/p/Cj4ToNRuRZY/?igshid=YmMyMTA2M2Y=

Kris is an AI artist I follow on Instagram.
She created her own AI model, training it to generate her own face. So now she can ask it to create an image of her and it comes out awesome. I’m really tempted to try it too, but I don’t have the hardware. I don’t think my laptop will be able to handle it.
Hmmm. I wonder if it’s something I could do with AWS. Curious. Curious. Like I need another project. Please, someone stop me.

But I’m linking to this specific post because she’s sharing the figures of exactly how much she makes from the images submitted to stock.

It’s actually pretty nice numbers. I’m sure it depends on how trending the images are, how much in demand, but still it makes me hopeful that maybe it’s worth the effort.

I should also add that Kris is kind of a big deal historically speaking. She holds the very first copyright for AI generated art which she got for the draft of her graphic novel. Currently, the novel is being finalized with the help of a digital artist.

There’s a whole story about that. I can tell you what I know if interested.

[Kamiccola]

I made the plunge and submitted 3 images to Shutterstock and Adobe stock.
One of them I manipulated to make it nicer but the other two went in as is.
Now we wait to see if they get approved.

There’s a big chance they won’t because they’re portraits, so even though they’re not photorealistic (they look like a cross between an oil painting and a digital painting), still they look like a model posed for the painting, so there’s a chance that they’ll get rejected. I likely have the best chances with the medusa one which is this fun image of a woman with snake scales on her face and a snake on her head. They were all attempts to create one of my characters but this is the only one that came out with cool scales like I wanted. Unfortunately, she doesn’t fit my character.

But if all three images get rejected by both stock platforms then I think it will be clearer that portraits are out the window and I should stick to abstract art or settings.

[J.L.O]

I like gyst stories…you know the antithesis to our writing styles. Lol

[Kamiccola]

I’ll take this as an invitation then. This is how much I’ve gathered about the situation.

Summary

She wrote the script for her story a while back intending to animate it on her own but, well, creating animation on your own is hard. She didn’t finish and put it aside.

Then AI generators became a thing so she started playing with them (Midjourney is her go to). She started creating scenes from her story and to have a consistent character, she used the actress Zandaya temporarily as a placeholder until she figured out how else to do it.

Soon enough, she created stills from her story the way she always imagined them and she shared online because you know, that’s what we do, we share, but it backfired on her. Someone reshared it without credit, and while it might have started innocently as “look at this cool thing,” it sparked a lot of debate and some criticism because she used the actress in it. And so it went viral without context, letting people create their own context. It didn’t matter that it was a proof of concept created with the intention of, hey, look how cool of a story it is. Imagine how cool it will be when I’m done. It was a bit late and her explanations didn’t matter.

Since it started going viral, she got worried that the situation was getting out of hand. So she filed for a copyright to protect her work. It was the first AI art to succeed at getting a copyright. It definitely is because as a whole, it’s a graphic novel. AI images alone I don’t think the courts would recognize (yet). But then that also got questionable publicity because it’s the first copyright of it’s kind and Zandaya look-alike is still in it.

She explained that it’s still a draft and she plans to copyright it again once it’s finalized. Her initial intent was to train her own AI (which she succeeded at) and use her own face in the final graphic novel. But well, the publicity worked out well for a change because one company offered to create an animated movie that used her AI art. Animating it had always been her dream so she jumped at the offer. She didn’t feel comfortable starring in it so they’re hiring an actress for that purpose.

And so now while the movie is in production, she’s working with an illustrator who’s tweaking the face of Zandaya to look more like the actress they hired. Per one of her messages, it looks like she’s planning to donate some (all?) proceeds from the novel to charity.

I only saw a few of the Zandaya pages and know nothing about the story but I’m really curious and want to watch that movie now.

So a lesson to learn from her experiences: be careful with using the likeness of real people. You don’t want your work to go viral for the wrong reasons.

[J.L.O]

Wattpad is full of face claims. This is stuff that NEEDS to be cautioned because a lot of us are doing at least mood boards, where we feel art is secondary to our writing.

I personally don’t think I’m any better in one or the other, the exception is that you can see where I’ve succeeded in art in general, and can easily see someone else is better at x,y,z while writing takes more effort on the readers to discern.

That and I don’t AI generate or fanfic in writing. The methodology is dynamically different. I’m basically repairing for “art”.

[J.L.O]

I’ve not really cared too much about sources for “free stories”.

[J.L.O]

Oh, “stained glass lovers” was the prompt for this one, with a base pic of a couple…

That Chameleon is so honking adorable.

Also: fair warning, Dream will make a topless man into a topless woman, and you can get a nude out of it from nowhere. To double-check the other end, I gave it a nude source, and it will take a nude source and work with it without censoring.

Now, these are clothed renders, but you can easily tell that the original was not…

So, visually safe, but mentally not.

[TheTigerWriter]

I got this thing using the free version of midjourney. Not doing any more because I’m not willing to spend money like that Although it does look cool, I did specifically say a burning castle at night time.