AI Generative Art

[Trike]

20 minutes ago, Techwright said:

 

This looks like some alternate universe cross between The Phantom and The Green Hornet, which I find rather cool.

 

By the way, what is "Art Station Style" that I see a couple of you mentioning?

No idea, but it somehow focuses the algorithm to be consistent. Sbloyd used it in a prompt so I tried it. Whenever I leave it out the results really go off the rails. I suspect it’s stealing from the artists on that site.

 

Which is why I’d never use AI-generated art for commercial purposes, but as an entertaining mini-game of “persuade the idiot to do what you want” I’m all for it.

 

As for The Phantom + The Green Hornet, there’s a good name: The Phantom Hornet - the ghost who flies. :)

[Damoklese]

[Trike]

Join the Space Patrol! See the Galaxy! Meet new friends!

 

 

Welcome to the Atomic Age of spaceflight! We have all the latest 1954 technology!
 

 

The farthest frontier has endless wonders and out-of-this-world encounters!

 

Learn technical skills that will stand you in good stead for a lifetime!

 

 

Multiple career paths available! From Alien Diplomacy all the way to Zoological Exploration!

 

 

Serve aboard the most technologically advanced spacecraft ever devised!

 

The Space Patrol stands ready to defend the galaxy!

 

Sign up with the Space Patrol! Adventure awaits! See your local recruiter today!

Edited February 4 by Trike
Typo

[BurtHutt]

Wow. I haven't messed around with the AI tools yet but wow...that's insane stuff it can produce.

 

For those of you that are making all of these cool pieces, how long does it take you to do one?

[sbloyd]

1 hour ago, BurtHutt said:

Wow. I haven't messed around with the AI tools yet but wow...that's insane stuff it can produce.

 

For those of you that are making all of these cool pieces, how long does it take you to do one?

 

Depends.  Using Microsoft's image creator, using their tokens, it's a couple seconds.  Once you run out of tokens, it's a few minutes.

[sbloyd]

6 hours ago, Techwright said:

 

This looks like some alternate universe cross between The Phantom and The Green Hornet, which I find rather cool.

 

By the way, what is "Art Station Style" that I see a couple of you mentioning?

 

It's just a way to make the system limit itself in certain ways.  A lot of the front-page artists on Artstation have some similar design themes, so it's a handy one I can remember.  You could also do "Style of John Byrne" and it transforms results to something like  and "cyberpunk style" gets me .

 

Not very different, but consistent.  Taking out the style requestor, I get these four results: 

...which, while not BAD, are inconsistent in style from result to result.  

 

At the same time, you can pull out "cel shaded" which forces a cartoon/comic book style and replace it with something like "painterly style" and get something akin to  which can be very appealing.  Sometimes it doesn't know what to do with some style prompts; I asked it for "Impressionistic style" and got  and  and  in the same round of outputs.  Really though, that wasn't fair, since Impressionism isn't something there's a lot of superhero picture samples around. A "watercolor painting" style prompt yields .  You can also try things like "lineart style" or you could go way retro with "style of roy lichtenstein" and get  type results.

 

[Trike]

13 minutes ago, Krimson said:

There's not a single Stable Diffusion Checkpoint model that contains a single pixel of art from any artist. 

I’m given to understand that Adobe’s Photoshop AI plugin Firefly was similarly trained on public domain art in order to avoid legal hassles. I’ve been primarily using Bing’s Image Creator, which often returns things that are suspiciously similar to existing art.

 

I specifically did my Space Patrol series to mimic government art of the 1940s/1950s while using the prompt “1930s pulp” to try and avoid copyright infringement. Not sure I was entirely successful, but I gave it a go.

[MagicalAct]

This was my first time messing around with AI stuff in general on kind of lazy afternoon. Had some good results, and some not so good results. Trying to figure out how to do my parameters better as I struggled with what I thought would be the easier characters and the more non linear ones seemed to come out well right away. We have Snow-Dragon, an Ice/MA dom, Draculad, a Mind/Savage dom, and Alpha Animal, a Beasts/Poison MM.

 

[Trike]

3 hours ago, BurtHutt said:

Wow. I haven't messed around with the AI tools yet but wow...that's insane stuff it can produce.

 

For those of you that are making all of these cool pieces, how long does it take you to do one?

 

1 hour ago, sbloyd said:

 

Depends.  Using Microsoft's image creator, using their tokens, it's a couple seconds.  Once you run out of tokens, it's a few minutes.

I’ve found it also depends on the day and time. Saturday morning around 10 am Eastern it gave me the message “too busy, try again later”. But then in the middle of the night last night between 1 am and 4 am Sunday I was able to generate a hundreds of those Space Patrol images fairly quickly despite not having any tokens.

[Trike]

12 minutes ago, MagicalAct said:

This was my first time messing around with AI stuff in general on kind of lazy afternoon. Had some good results, and some not so good results. Trying to figure out how to do my parameters better as I struggled with what I thought would be the easier characters and the more non linear ones seemed to come out well right away. We have Snow-Dragon, an Ice/MA dom, Draculad, a Mind/Savage dom, and Alpha Animal, a Beasts/Poison MM.

 

Those look great. And I have to compliment you on the name “Draculad”. Brilliant! Wish I’d thought of it.
 

Did you use Bing for these?

 

I’ve found that it has definite blind spots when it comes to some things. I’ve given up trying to recreate many characters because it was just returning totally random stuff that was completely unrelated to what I’d asked for. For instance, see those short aliens in the swamp in the Space Patrol pics? I initially asked for them to look like a platypus. It gave me the most bizarre creatures in response. The most hilarious ones were Kermit the Frog-looking things, and it kept trying to give me alligators — even for the human astronauts! 😂 I eventually asked for purple and teal otters, which it liked better, but still rendered weird stuff.

 

These are all from its idea of “platypus”. 😜

 

 

 

[Trike]

4 hours ago, BurtHutt said:

Wow. I haven't messed around with the AI tools yet but wow...that's insane stuff it can produce.

 

For those of you that are making all of these cool pieces, how long does it take you to do one?

I recorded how long it takes. This is pretty typical. Added a bit of music for fun.

 

https://www.instagram.com/reel/C28k2oUO2rG/?igsh=azZ4cjlicTNwN3Aw

 

The prompt: “art station style, a steampunk duck in a dungeon, spraying robed cultists with water”

 

The results are AWESOME.

 

Edit: BTW, you can see I have zero tokens. The indicator is at the end of box where you type in the prompt. Having tokens prioritizes your render, I guess, and they seem to refill every day.

 

Edited February 5 by Trike

[sbloyd]

Yeah, AI generative art doesn't *directly* take anything from anyone.  It does , for lack of a better word, learn by examining thousands of pieces of art to create algorithmic data which is associated with certain keywords, and which gets used as part of the output when we use various keywords in our prompt requests.  The question, which is beyond the scope of discussion here, is whether this is ethical or not.  I have a fine art degree equivalent to a Masters, and I'll tell you, we are made to mimic other artists' work over and over; you have to ask yourself whether there's any difference between the two. 

 

Me, I see AI generative art as a toy, and a tool, to work on my own art.  One of the software packages I use includes built-in, posable 3D models of people, common locations, et cetera.  You can literally load them into the drawing program, pose them, and trace them.  It's just another tool.

 

One nice thing Bing has done is an attempt to exclude using specific peoples' likenesses - generally it will kick out a prompt with someone's name - though from time to time people have figured out ways around it.  Those situations - from MS's generator to Stable Diffusion to other generators - are where we get weird and disturbing things involving deepfakes.

 

Anyhow, it's up to each of us individually to decide how we use these tools.  If you feel like it's inappropriate... then don't use 'em.  I certainly don't try to pass off the results as my own work, but it's very interesting to generate a dozen or so images and see how the different weights and randomizers and seeds all fall together to create an image. Either way, I hope we can agree to let it rest (at least here - this doesn't feel like the right venue for such discussions).

[sbloyd]

Back to the fun stuff!

 

One interesting thing about Stable Diffusion is that you can feed it a base image alongside a text prompt.  I took a screenshot of my main, Horizon Twilight, and did a couple iterations of generation on it:

 

 

...might have to go visit Icon, I think!

(And hoo-boy, my Stable Diffusion likes them thicc!)

[Marshal_General]

32 minutes ago, sbloyd said:

Back to the fun stuff!

 

One interesting thing about Stable Diffusion is that you can feed it a base image alongside a text prompt.  I took a screenshot of my main, Horizon Twilight, and did a couple iterations of generation on it:

 

 

...might have to go visit Icon, I think!

(And hoo-boy, my Stable Diffusion likes them thicc!)

I see her hips have been keeping up with the Kardashians.

[Dynamo-Joe]

1 hour ago, sbloyd said:

Back to the fun stuff!

 

One interesting thing about Stable Diffusion is that you can feed it a base image alongside a text prompt.  I took a screenshot of my main, Horizon Twilight, and did a couple iterations of generation on it:

 

 

...might have to go visit Icon, I think!

(And hoo-boy, my Stable Diffusion likes them thicc!)

 

Dang! She's got that Pixar Mom booty going on! 😅

This is not a bad thing. 😁

[MagicalAct]

8 hours ago, Trike said:

Those look great. And I have to compliment you on the name “Draculad”. Brilliant! Wish I’d thought of it.
 

Did you use Bing for these?

 

I’ve found that it has definite blind spots when it comes to some things. I’ve given up trying to recreate many characters because it was just returning totally random stuff that was completely unrelated to what I’d asked for. For instance, see those short aliens in the swamp in the Space Patrol pics? I initially asked for them to look like a platypus. It gave me the most bizarre creatures in response. The most hilarious ones were Kermit the Frog-looking things, and it kept trying to give me alligators — even for the human astronauts! 😂 I eventually asked for purple and teal otters, which it liked better, but still rendered weird stuff.

 

These are all from its idea of “platypus”. 😜

 

 

 

Thank you kindly, a lot of people seem to be fond of Draculad lol. And I used civitai that someone had mentioned earlier. I can't get it to go for certain styles as far as I found outside of the different model filters but it seems pretty easy to use.

[MagicalAct]

My power has been out all day, so I've just been messing with this, trying the different model styles. We have another crack at Alpha that came shockingly close to his in game look, Son of Surtur, my Fire/Earth Dom, and Masked Hornet, my Poison/Elec defender. 

Edited February 5 by MagicalAct
Had to make it cleaner to read.

[choake]

14 hours ago, Trike said:

I recorded how long it takes. This is pretty typical. Added a bit of music for fun.

 

https://www.instagram.com/reel/C28k2oUO2rG/?igsh=azZ4cjlicTNwN3Aw

 

The prompt: “art station style, a steampunk duck in a dungeon, spraying robed cultists with water”

 

The results are AWESOME.

 

Edit: BTW, you can see I have zero tokens. The indicator is at the end of box where you type in the prompt. Having tokens prioritizes your render, I guess, and they seem to refill every day.

 

LOLOLOL.. thank you!! this is fantastic!

[sbloyd]

RE: my character being a total dump truck... I think I actually did have a Pixar preset flagged, hah hah!

[Trike]

On 2/4/2024 at 10:51 AM, Techwright said:

 

This looks like some alternate universe cross between The Phantom and The Green Hornet, which I find rather cool.

 

This was the first thing I saw when I logged in last night. 😂

 

 

[ThaOGDreamWeaver]

Oh look, I'm a collectible.

 

 

 

 

...and here's the MS template for it.

(Tip: it doesn't like the word superheroine, because it contains a banned word related to the drug. Clearly it ain't AI about everything. Specify Superhero, female.)

 

https://designer.microsoft.com/image-creator?p=Funko+figure+of+[an+activity+or+role]%2C+[female+or+male]%2C+called+[Name]%2C+wearing+a+[specific+clothes+and+glasses]%2C+[specific+hair]+and+has+[accessories].+Holding+a+[something].+The+Funko+is+displayed+inside+a+Funko+box+with+[text]+text+and+[logo]+logo+for+the+box%2C+allowing+visibility+of+the+figure%2C+typography%2C+3D+render

[Trike]

Do you prefer Fantasy or Science Fiction? Tavern or bar?

 

[Techwright]

I turned my attention to my MA/SR scrapper Ace Barnstormer, and although I was never able to get the A.I. to completely understand me, it did deliver some cool results, even if it did feel so insecure that it had to supply two sets of goggles to some pilots.

 

[ROGUEdenied]

Sometimes you gotta bring your extra nightvision goggles too

 

[sbloyd]

Yo, dawg, we heard you liked goggles...