🤯 Any other AI art addicts here? 🤯

[CoffeebyNight]

Haha, no problems about it being wordy! It was an interesting read! I disagree with some of this - namely the “I’d have to be more exposed to it” bit. I’m constantly, constantly bombarded with it. I’ve seen what actual artists who turned to AI are creating, and it’s still clear as day. Quality does vary across the board, but I see it every single day, namely from artists. I lurk here sometimes and see what the current discussions are about - every piece posted here is obvious (not saying that’s a bad thing at all, btw, AI sheen isn’t something you can get rid of atm).

We both agree that AI won’t replace artists (or illustrators, imo). I also model, and I don’t see a world where people ditch models either.

@Kamiccola

It’s a derivative work, not a copy.

This is super interesting to me, and I would LOVE to hear your thoughts on why you think this? It’s demonstratably false that it can’t replicate things exactly. Midjourney, even without names of IPs in the prompts, has been spitting out exact 1:1 replicas (except with the AI sheen) of Marvel movies, Batman etc. Many generators are, still to this day, spitting out 1:1 replicas of watermarks of artists and stock sites. I am currently taking a break from Twitter because too much AI lol, but I can grab you some links if you want to take a mooch?

All in all, it’s really, really nice to hear your thoughts Thank you for sharing. I disagree with some bits, and agree with others! I’m also not worried about AI books on their own - they are, indeed, trash. I’m not sure if I deleted it out of my last message, but I feel safe in my jobs. Sure, I took a hit to begin with, but things seem back to normalcy. Other than parts of the internet being harder to use now, people undercutting my prices because they can generate 100 different files in the time it takes me to make a draft.

Seems we also agree that it’s namely the beginners who will be in trouble financial. I like to think I’m secure enough in every bit of work I do (except for a hiatus from modelling) to keep getting work, and it seems the newer artists etc are carrying on. It hurts to think that they might eventually give up because they can’t crack into their chosen industries, though

I would still love to hear your ideas on how we can metaphorically put the genie back into the bottle RE the people using gAI to blackmail, scam and commit SA online, if you feel down to discuss that. If not, that’s totally fine! And again just as reassurance, I REALLY hope I’m not coming off as hostile here I believe gAI can be insanely useful in the right hands and I hope for a future where it makes it into those right hands!

[Kamiccola]

You’re definitely not hostile. It’s refreshing to see a civilized conversation.

There are some polarizing views out there that make it difficult to exist. This whole topic is not black and white. There is a lot of nuance that can’t be expressed on social media. Character limit hurts us.

Regarding the replicas and accuracy, I think it has to do with the number of images available. AI can recreate the Mona Lisa pretty well because this image has millions of iterations on the web. Bigger likelihood of unintentional replica.

I think it’s a paradoxical limitation of the tool. You don’t want it to copy things exactly but you want it to be accurate in what it depicts and not give you 7 fingers. The more accurate it gets, the less original it is.
And I don’t think there’s a way to prevent it other than blacklisting certain words (which it does already - excessively).

The algorithm itself doesn’t store the images so it’s not like it can notice, oh no, that’s too close to original, let me change it up.

I agree that it’s a problem in the wrong hands. It’s easier now to create fake content and there will be people who fall for it, thinking it’s real. Common sense is the answer but I think we all know that some people don’t have any - especially once emotions get involved.

I don’t know what the solution is. There should be a way for AI generators to insert a signature in their images to make it possible for a specialized application to detect it. But there are all these independent models that anyone with the right hardware can create now. You can’t regulate what people brew on their own.

@CoffeebyNight

I disagree with some of this - namely the “I’d have to be more exposed to it” bit. I’m constantly, constantly bombarded with it.

Oh, I’m sure you get to see it. No one can escape it now. I have to tack on -ai in every image search nowadays. So much crappy content. When will Google stop indexing these sites?

I meant more the exploration of the AI spaces, hanging out in places where everyone creates and shares, and of course, trying out the tools on your own to see its possibilities and limitations. People who are against it normally don’t. But I admit I have no clue where you do or don’t go, so I can’t speak on your experiences.

---

Okay, so the elephant in the room.
Art theft.

That’s a tricky one because it’s not a clear area.

The technology doesn’t copy images and doesn’t store them anywhere. It’s a set of algorithms that calculate things based on the 1s and 0s given to it. I don’t see anything immoral here. I think it’s pretty amazing, really.

The obtaining of the images to create those 1s and 0s is the gray area.
I don’t think it was illegal, it wasn’t theft, but I think that there was an opportunity to do it in a way that rewarded the contributions. An opportunity that wasn’t taken.

Arts have already been so poorly valued.
This amazing technology that’s a stepping stone for the future of the creative movement could have rewarded the artists that helped make it possible. But corporate greed won.
Artists are still being poorly valued.

I think it’s a symptom of how easily our society takes arts for granted. It’s a larger issue, not just AI related. AI has just become the stressor that pointed out our society’s flaws.

So I think there’s a valid argument to start this over and do it right. I just don’t have much hope that it will happen with the current AI models.

Artists creating their own model would have been an awesome solution. I would totally prefer to spend my money there. I’d pay for that Kickstarter.

[MatthewJH]

^Too many words, not enough pictures.

I don’t read the articles, I only look at the pictures.

Ahem. Here are more “old MJ prompt meet SDXL” experiments.

Yasmine, when she was Zorik's girlfriend and protegee.

I sometimes think the Bedouin mercenaries in my stories are not intimidating enough, compared to the Bedouin in the IDF. But Yasmine, the only non-Special Forces qualified member of their group, did resolve any doubts about her value or courage as a mercenary by spear-tackling a SEAL. Later she materialized beside a SAS (first novel) and a Sayeret (second novel) sniper team, greeting them with words to the effect of “You’re lucky I’m on your side”.

[MatthewJH]

Some more simplistic early MJ prompt experiments in SDXL. No loras, negative prompts, or post-production edits.

Interesting reflection and lighting effects

^The drunk fling lovechild of a HSV and a Mustang…me wants…

^Fish-eye lens effect?

Cars with holographic paint

[Kamiccola]

Yo

Midjourney v6: 15 Styles Inspired by Different Techniques

I need to try these.
My credits just renewed. Looks like I’m about to spend them all.

[MatthewJH]

Looks like AI images are seeping into marketing. Following examples discovered by others on Reddit.
A Cars Australia website.
A high-end hotel website. Note the bikini girls’ feet (in the scrolling image advertisement section).
The term “actual results may vary” is relevant once again. Those images are average by my knucklehead prompting standards, but they seem good enough for someone out there.
On this subject, here’s a new feature in Mid Journey.

Why Is Nobody Talking About These AI Use Cases?

Update: And this just appeared in my subscriptions feed.

Create Consistent Personalized AI Characters 𝙄𝙉𝙎𝙏𝘼𝙉𝙏𝙇𝙔! Open Source Stylization Method!

[J.L.O]

The only thing that would really be useful is if they get behind the idea of allowing AI to forget, even if it’s a toggled switch that turns off it’s “memory”.

Teaching AI how to forget at scale

This would make ethical easier to “access”. If it trained on your art, but is being used specifically for “not your art” projects, then I could come into a program and retrain it to my stuff, not have to start from scratch. From the deep studies I’ve seen individual artists work on (just in passing), they say it really takes about 10 pieces to start churning out “your style”, and if yours was 1 mil images back, it should be old and forgotten, like this.

I’ve been a photomanipulator and have drawn since forever, but I don’t bother to call myself an artist. The way I see it is it should be more for free use situations. If I’m ripping off Riker from STTNG for a meme, then memes should be held to the same level of theft accusation. I certainly don’t own Paramount’s work.

I spend as much time editing AI as I do any other image source, so it’s not really a time saver to me, so I don’t find it any more or less convenient to use.

[MatthewJH]

Playing with other early, early MJ prompts.

Zorik, when he first met Karen and became her bodyguard (and closest friend)

^Good to see SDXL associates the Golan Heights with wine.
L’Chaim to all.

Either SDXL or my prompt writing struggles to depict Zorik’s size. Zorik is introduced, and often refereed to, as a literal giant. 7’4" tall, about 4’5" across his shoulders, barrel chest, and has enough strength to carry and fire a FN-MAG like a M-16…and sprint* through a Romanian winter forest in full alpine gear, with a FN-FAL and a 75L Berghaus pack…

*No one that big should move that fast ~ Adeptus Astartes quote (or a quote about Halo’s Spartans).

[MatthewJH]

Found this on r/singularity.

Sister of Battle - Warhammer 40k (Unofficial)

The art and animation are AI generated (and maybe the sound too). Only took 60 hours of work from storyboard drafts to final product. Apparently the creator used midjourney, runwayml, magnific AI, and a lot of prompting credits (I’m guessing the last bit).

Some MJ prompts meet SDXL images / Xmas stocking fillers.

^Nice attempt with the disc brakes.

[Kamiccola]

Say what you want about the AI quality of some of these clips, as a whole, it’s pretty awesome. I was most impressed by the dress made of scrolls. So cool.

I’m at work so I watched it without sound, but it’s like I could hear the music anyway.

It really offers a glimpse of the future of entertainment. I hadn’t given it a lot of thought before how generative AI can be used in filmmaking, but I can see it as an alternative to CGI. It’s faster, cheaper, more imaginative. You need to reshoot a scene for the final director’s cut? No need. AI can fill in the gaps or fix mistakes. Actor got pregnant during a long season? No problem. Just remove the baby bump with AI. Kid actors growing up too fast? No problem. How old do you want them to look?

I imagine movies and TV in a few years will be using it a lot, and just like CGI, there will be good AI movies and bad AI movies.

[Kamiccola]

Trying out some new prompts for a book cover.

I’ve always noticed that the results are more interesting when you don’t try to create a cover-ready image but allow AI to show you a book.
I’ve found it frustrating before, but I think I finally understand why this happens.

The images that are more suited to use as a cover are trying to mimic the premade covers while the ones that look like books mimic pictures of published books.

That’s where the quality difference comes from!

Anyway, so I’m looking at these and I’m inspired. I wonder, can I draw something like that?

Summary

That last one is my fave. The 3d texture looks so awesome.

I also have started wondering about this cool possibility. Check this out.

To get a book embossed in gold such as these would be very costly. I’d need to work with a specialized printer, assuming that those printers even work with Indies. I don’t expect people to want to spend $30 for a book by an unknown author just because it had a cool cover.

But what if I printed the copies of ordinary quality covers and then put the gold embossing on it myself?
Imagine this. A handmade book cover at a reasonable price. That’s like limited edition collectible right there. Maybe a few people would buy it if only for the cool factor.

As to how.
There’s this artist I’ve been following on Instagram that uses gold foil. And I’ve seen that foil in freaking Target! It’s not expensive.

Example of the foil in action.

Mallarie Bundy on Instagram: "They came out so much better than I expect....

I would just need to experiment with my technique to learn how to do it well and experiment with the type of book coating that would be ideal for that. I could get some bargain books as my test subjects to check how well the gold sticks.

As is, I have metalic gold markers and gel pens. The pens could be tricky on a cover but the marker might work. I just don’t know how well it would stay put. How durable it would be. Or if I’d need to put a finishing varnish or something like that. Another thing to test, I guess.

[MatthewJH]

Am currently pillaging the Civit.ai website in search of more photo-realistic™ Checkpoints to play with. Almost out of hard drive space. May need a larger SSD (+4TB). Who knew generative AI could be such an expensive hobby.

Anyway. The prototypes for generative-fill AI, for 3D renders, have been around for a while. NVIDIA called their version DLSS. Photoshop now has a generative-fill AI system too.

Photoshop Generative Fill - 20 EPIC Uses, SUPERFAST!

Top 7 NEW Features Explained! - Photoshop 2024

And a Polish advertising company is now experimenting with TV ads created with AI software.

Maxon Forte - Trener uwodzenia

Far from perfect, or passable, but it’s early days.

PS. In reference to my previous post, a veteran albino Sisters of Battle* bodyguard has been added to my 2026 Christmas list. I should have found a winning Power Ball ticket by then…maybe AI can help there too (not financial advice)…

*I couldn't find the desired Sisters of Battle artwork on internet image search, so I created some stand-ins with SDXL.

My “roll forehead across keyboard” style of prompting seems to work well.

^Wonder if generative-fill AI can remove that bright red patch on her hair.

^Who introduced Ashley to cosplay?

*Update: Found it.

Some women are more dangerous than shotguns ~ The Godfather film.

[MatthewJH]

Random thought. How about using gold-leaf (gold foil) with a thin laminate instead of embossed gold for future book covers. Might be cheaper, easier, more durable. Or maybe only use the real gold decorations for limited publication runs.

Just found this video, bound to be more gold-leafing tips around the internet.

Decorating a Leather Book Cover with Gold Foil: "Traditional" Methods on a DIY Budget

[MatthewJH]

Decided to experiment with the Sisters of Battle lora. Looks promising.

And the Space Marine armor lora has some potential.

[Kamiccola]

Just having fun.

MJ V 5.2

Vs V 6

I like the composition of 5.2 better but 6 is better at getting more accurate with the prompt. Plus you can actually see the details of the fiber. Cool.

Okay, let’s get weirder.

Weird = fun

FYI, the last one is a flea.

I absolutely love how it chooses different style of glasses for different creatures. There’s a stereotype at work here that AI noticed that maybe we wouldn’t.

[MatthewJH]

…Resist urge to restart MJ subscription, resist urge to restart MJ subscription, resist urge to restart MJ subscription…

How do those images look up-scaled to 2048x2048? I’m considering going back to MJ, updating old favourite MJ prompts, and up-scaling the v6 images with SDXL for a comparison exercise.

Oh, a possible theme song for your Weird gallery…

Will Smith - Men In Black (Video Version)

[Kamiccola]

Playing with v6 some more by generating Sam and Ian again.

So far my observations:

1. Better results with --stylize 500. Raw is usually a bit disappointing.

2. Most images are of good quality, it adheres to the prompt well, but there are strong biases.

My biggest frustration so far is that it keeps drawing Sam as a doll. You know what I mean, big eyes, nonexistent nose, perky cleavage.

As if that isn’t bad enough, dressing her in the emerald blouse that she wears in the scene I was trying to recreate 99% results in the an itty bitty frilly thing that barely covers her chest. Changing the prompt, adding modest, etc doesn’t seem to help. I had to put her back in the flannel shirt again. Hmmm. I haven’t tried emerald flannel. I’ll give that a go.

I really hope that it won’t become a trend in v 6, these doll-like characters. That would suck.

Anyway, here are my favorites so far.

Update. Emerald flannel doesn’t fix the issues. In fact, I can’t seem to be able to produce anything decent today. Nothing close to the output last night.

Instead, now it’s making Ian shirtless and ripped. I like abs as any other gal but that’s not what I’m trying to create.

[MatthewJH]

Some AI image generation news. Might be of interest.

Midjourney Alpha, Krea LIVE AI and NEXT LEVEL AI Art

[Kamiccola]

Yo, MJ v6 can do double exposure.

Anyway, I wanted to share the latest gem which MJ gave me.
It combined Samantha and Ian into one person and… I can‘t help but think they’re kinda hot.

I gave up for now on trying to create that gazebo scene. Something about mentioning gazebo triggers a bias that dresses Sam into a skimpy little shirt, exposing her belly (the kind you see above but skimpier). I’ve had enough of that.

This one is my favorite today. I’m tempted to draw them. What are the chances it will come out well?

I should stop creating them over and over… sigh. Or maybe not.

[Kamiccola]

I checked out Alpha. Looks cool. Maybe I’ll start creating there instead of on my server. (I think I racked up enough to go over 5000 back when I had the higher plan and it was about to expire - I did some massive generating on relax mode, basically using Describe on hundreds of images from my laptop, lol, good times).

The only reason to stay on my server for now is that I have everything organized in channels and threads under those channels.

There are new folder options in MJ now though that I haven’t played with yet. Maybe that will solve the pain point of organization but it will be an extra step. Though, using discord is an extra step too so maybe it will even out.

There are some cool options in discord though like creating your own styles, inpainting, etc. I wonder when those will be available on the Alpha.