AI Generative Art

[Trike]

15 hours ago, twozerofoxtrot said:

 

This dude goes bowling a lot!!

Is that what the kids call it these days? In my youth it was known as “pocket pool”.

[Trike]

23 hours ago, Techwright said:

 

You've identified the problem I've been having.  I don't know the real words, if there are any, for a variety of costume parts from the game.  For example, my retro-future space ranger, Major Ray Gunn, uses what the game calls "fury wings" along the side of his helmet, but this apparently is not a real world term for that part, if it even has one.

Did you try “winged helmet” or “Norse helm” and such? Sometimes it has it filed under odd things like that.

 

I had an impossible time getting it to do jetpacks, so I googled the word and I tried all the alternatives, getting nothing. It’s weird what blank spaces it has.

 

I was on a Sinbad kick the other day and tried to get Bing to do fights between Sinbad & Kali and Sinbad & Cyclops. And BOY HOWDY does it have a terrifying idea about what a cyclops looks like. Trying to redefine it just made it all weirder. Even ended up with a triclops or two.

[The Tsaurian]

On 1/25/2024 at 8:26 AM, sbloyd said:

At work I've been messing around with the Image Creator in Microsoft's Bing, since I don't have access to my Stable Diffusion home setup.  I've been very pleased with the results I've been getting.

 

The core prompt I've been using is

cel shaded, style of artstation, city of heroes, background constrained in a circle, superhero

followed by more specific instructions.

Tweaked it a bit for my Tanker, but that prompt is CA$H MONEY!

 

cel shaded, style of artstation, city of villains, background constrained in a circle, supervillain, NO CAPE, monstrous reptilian humanoid, blue teal scales, crystal spikes grow from back and shoulders, mohawk of horns, rusted chains drape over chest and shoulders, roaring a challenge

 

yielded

 

for reference, my Tanker

Edited February 18 by The Tsaurian

[Trike]

On 2/16/2024 at 6:20 PM, Trike said:

On 2/15/2024 at 7:11 PM, Techwright said:

You've identified the problem I've been having.  I don't know the real words, if there are any, for a variety of costume parts from the game.  For example, my retro-future space ranger, Major Ray Gunn, uses what the game calls "fury wings" along the side of his helmet, but this apparently is not a real world term for that part, if it even has one.

Did you try “winged helmet” or “Norse helm” and such? Sometimes it has it filed under odd things like that

I gave various things a try and Bing no likey. 
 

Here are the ones that came the closest:

 

 

The various prompts, all of which resulted in winged backpacks most of the time. There were a couple more but I neglected to copy them. Mostly of the same ilk, and you can see the trends. I didn’t try “Space Ranger” because from previous experience it gives me ersatz Power Rangers.

 

art station style, photo realist, golden age comics Space hero, laser pistol, open face helmet with stylized wing earpieces

 

art station style, cel shaded, golden age comics Space hero, laser pistol, open face helmet with stylized wing earpieces
 

art station style, cel shaded, 1920s Space Adventurer, laser pistol, winged space viking helmet

 

art station style, cel shaded, science fiction, 1940s Space Adventurer, laser pistol, winged space viking helmet

 

 

 

[Trike]

I tried several times to do an AI version of my mutant Mr. Magma, a guy who became a living volcano when he was assisting some scientists on a tropical island. (Sorry, currently 850 miles away from my computer so I don’t have the screenshot of his bio.) I asked for his skin to be basalt and lava and the program nailed that, but for some reason it kept giving me giant versions of him. I tried all sorts of qualifiers like giving an exact height (5’7”) or saying “average sized” or “short Latino man”, to no avail. I did manage to get a couple that were decent, but the enormous ones kept appearing. When I asked for the specific sizes they got either weirder or bigger.

 

This is actually pretty close to the game and what I wanted:

 

As soon as I requested more people in the scene, he started growing:

 

And then there were weird, random renders:

[Braddack]

Weather news, the best Superhero in Weather

 

 

https://ibb.co/1GNymzm

[choake]

Totally down for critiques and comments and all the good stuffs... It's been pretty awesome working into this style:

 

I've been working on refiing my Nub guide around the towns, and i think from my typical super fancy stuffs i sell and pimp out, getting to this place was hella hard but totally worth it.. All the text and words and layouts are generated auto.. only my coding and story arc -- Which was MadNESS to get.. 

 

My first few goes at working it out..

After my main cover art types above... 

 

And then..... Early today....

[Techwright]

5 hours ago, choake said:

After my main cover art types above... 

 

And then..... Early today....

 

 

Of the two shots I selected above, the first is something I'd very much like to see added to the game, perhaps in a revitalized part of the city.  The latter shot, the statue looks more like an Achilles than an Atlas, but I find the look quite good.

 

[OldManMercy]

First attempt at Detrius, my living garbage br00t. Prompts: Large humanoid made of decaying organic matter and garbage 3d CGI

[Techwright]

Decided to turn my attention to the concept of hero statues throughout a city.  Thematically, I looked at five sub-concepts:

 

1. A monument similar to the Soldiers & Sailors monument in Indianapolis.  If you're not familiar with it, the S&S monument is to Indy what the Gateway Arch is to St. Louis. 

2. Random statues throughout a major city.

3. The concept of a memorial garden for the fallen.

4. What might have been the possibilities if Atlas were memorialized at the bridge he guarded in his final battle.

5. What a hero statue might look like at a lighthouse (Siren's Call)

 

The fight this time was with the Bing AI interpreting "superhero" to always mean "Superman" (or occasionally Spider-Man, Captain America, and Iron Man.  Oddly Batman wasn't a problem.)  Trying to get that S-in-a-diamond look to go away was a mess, and in the end, I acquiesced to it being in one of the shots.

 

 

[Techwright]

34 minutes ago, OldManMercy said:

First attempt at Detrius, my living garbage br00t. Prompts: Large humanoid made of decaying organic matter and garbage 3d CGI

 

 

Concept-wise, he sounds similar to Junkpile of X-men 2099 comics, though I've not seen that character in 30 years, and what remains online is a lot cleaner in look than I remember.

[choake]

 

 

3 hours ago, Techwright said:

Of the two shots I selected above, the first is something I'd very much like to see added to the game, perhaps in a revitalized part of the city.  The latter shot, the statue looks more like an Achilles than an Atlas, but I find the look quite good.

 

Thanks mate.. the 2nd group was the 'finally sorting out how to position and call up text and such' 

I have so many epic comic setup types now.. it's unreal..
 I have a few more that are proper atlas I got the turn out pretty well too.. And 100% agree with the ingame Layout -- In the Pocket D around the Truck Changer guy, NuLL the GuLL can take you to other realms of seasons and stuffs AP Seasons  <-- links are coh-wiki .:strangerDangers and rando links lol:.

 

 

 

[choake]

 

 

 

 

 

I mean LOL... This one is PerFection! -- Only a few small type fixes needed and it'll be awesome

 

 

Edited February 19 by choake
dbl img DeRp

[Trike]

28 minutes ago, choake said:

 

This is my favorite of all the ones you’ve done. It really looks like Atls Paark or Taos Isand. 😛

 

Seriously, though, drop in some actual text over that and it would be a terrific poster. It’s so,weird how the AI can’t spell, yet keeps trying to.

[choake]

4 hours ago, Trike said:

This is my favorite of all the ones you’ve done. It really looks like Atls Paark or Taos Isand. 😛

 

Seriously, though, drop in some actual text over that and it would be a terrific poster. It’s so,weird how the AI can’t spell, yet keeps trying to.

---

TY again for the input!! ||  🦄 🖖 I Bla blah Enjoy.. I like a challenge, and a different eye | 

 

 

Spoiler

 

 

---

YT Art Chan   BeHance | My Etsy

.:Dork Momet Over:. 

---

The Text is pretty simple to isolate and shift about, just don't really care about the vanilla pieces until I refine them later so im not too worries a out that..

ContentAwareTool FTW lol and Puppet Wrap ❤️ but loving the shift into This style

 

So... I've gotten to a pretty Decent repeated style.. and MAN getting the typography and placements / layouts was INSANE!! lol 

you  can see the tiny steps of of tweeks, but yeah.. almost there... 

 

p.S. I love being a TechnoMancer man.. This has been the most epic adventure over the last few years

 

Sorting out how to stage the spaces , remove the spaces to not gen in... was super complex but then its like PooF.. Finally worded my knowledge paramaters lol..

 

This was after sorting out the aobves earlier..

 

Negative Prompts FTW!! btw

---

 

THIS Below.. The slow progress and LvL up I get to look at in live time is such a great rush.. 

 

Tl;dR techno rambles removed lol

Love bringing out the chaos in the logic and then WiN -- .:doRk:.

.: my gpt assisstant bot I work on it's about as far as i can push it.. - consistent characters and such, it's just getting to that space that's RAWR.. 

 

Mapping out this space was the most difficult.. so many odd bits kept showing up lol

 

 

then

this was a fun jump and then im like Angle it I QUIT lol

and then the transition kinda spot of this still awesome 

Figured try it from above, much 'easier' to limit... 

 

 

And then PooF Again lol.. 

 

I'm still trying to sort out How not to do this or that.. But I think i'm doing pretty killer lol

 

❤️

Edited February 19 by choake
Rambles

[choake]

1 hour ago, Trike said:

This is my favorite of all the ones you’ve done. It really looks like Atls Paark or Taos Isand. 😛

 

Seriously, though, drop in some actual text over that and it would be a terrific poster. It’s so,weird how the AI can’t spell, yet keeps trying to.

I accept your Challenge to Sort of the Paragon with my PhotoShop wizardry.. Lets see.. 10:52 est.. is when i start.. Lets see what I can do quick 🙂

[choake]

12:08 est.. Took a while to play around but yeah.. Not to shabby less than 30mins

Not a aFlex or whatever, but in 15mins I had it fixes and then wanted to mess around with the people.. under an hour for 4 isolated people and words and just.. not bad lol

---

 

 

---

---

---

 

Edited February 19 by choake
derp

[sbloyd]

I'm glad sharing our prompts has helped people, that was part of what I was hoping when I started a thread specifically for AI generative art separate from the other art thread.

 

I was experimenting with my "make coloring sheets" AI prompt, which is really just my usual prompt with "line art, black and white" added, when I got the idea to include "screen tone" in it for shading.  After generating a handful of coloring sheets I wanted to see what just adding "screen tone" to my normal prompt, and I got this result for my Grav/Kin Counterweight:

city of heroes, background constrained in a circle, wide angle, full body, superhero, style of artstation,, korean woman, yellow and charcoal grey uniform, infinity symbol insignia on chest, yellow sunglasses, charcoal grey leather bolero jacket, cupping a swirl of energy in her hands, pixie haircut, hovering in flight, screen tone, no cape

 

Very moody.  I like it.

 

For those interested, the "coloring book" results looked like...

 

[battlewraith]

It looks like most of the people here using Microsoft's (?) AI generator. These results look pretty good, honestly a lot better than I expected.

I've been subscribed to Midjourney for almost a year now. There are some features that are pretty essential to me in that application. I don't know which of these features are part of other applications, but it would be a good idea to look for them.

 

1. Image weights. So when I'm looking for a certain kind of result in MJ, usually I link images at the start of the prompt to help guide it towards what I want. So if I wanted something like Batman exploring a dark warehouse, I might have a link to an image of Batman and a link to an image of a dark warehouse in the style I like. The AI will then take cues from the linked images and the prompt. MJ uses tags at the end of prompts to vary parameters. In this case there is an --iw (image weight) tag that tells the prompt how heavily to weight the influence of the linked images. The tag is capped at 2 to prevent people from making revenge porn, etc. from images of real people.

 

So that prompt might look like: /imagine [picture of Batman] [picture of dark warehouse] Batman exploring a dark warehouse, pen and ink comic book style illustration --c 35 --iw 2

 

"--c 35" stands for chaos 35. The purpose is to guarantee some variety in the output so that you don't get 4 super similar results when it processes the prompt. I think it goes up to a hundred but most people use something around 35.

 

2. Style ref. This is a new feature. It's very similar to image weights, but it's meant to take direction from the formal characteristics of an image but not the actual content of the linked image. So if I use an image weight of Batman and the picture has a coffee cup in the background, coffee cups will probably start showing up in my generations. If I use a text prompt that says "Batman in the style of John Singer Sargent", it will give me Batman in something like that style--but it will also start to throw in content related to Sargent's work. You might get a picture of Batman with a period hat for example. If I link an image of one of Sargent's paintings as a style ref, in theory it should only give me the influence of Sargent's style without bringing in associated content.  The --sw tag tells the prompt how much weight to give the sref, all the way up to 1000. You can also use multiple images in the style ref (e.g. Norman Rockwell, Jack Kirby, H.R. Giger all in one ref).

 

3. Inpainting. MJ has a function called "vary region." So in many cases you do a prompt and it gives you a result that's great but there is one small detail that's wrong. Or a limb is posed the wrong way or something. You do a vary region, which allows you to select the problem area and then do the command. You get a text box that allows you to add or change information relating to the original prompt (only in the selected area, it doesn't change anything else). It's not perfect, and might take several tries to get a good result but people do routinely use it to fix things in their generations.

 

I'm not trying to be a shill for Midjourney here. I think these are going to eventually standard features of any AI image generator. If your using something else, or are just new to this type of software, check to see if the application you're using has a version of these commands.

[sbloyd]

No problems!  I just use Microsoft's generator because it's not blocked at work, and I frequently have a lot of downtime while my students are working on projects.  I do frequently miss the more in-depth controls I have access to on my home machine's Stability install.

[Trike]

I downloaded Easy Diffusion and it really messed up my PC. Couldn’t even play CoH without crashing. Had to uninstall and defrag to get everything working properly again.

[sbloyd]

39 minutes ago, Trike said:

I downloaded Easy Diffusion and it really messed up my PC. Couldn’t even play CoH without crashing. Had to uninstall and defrag to get everything working properly again.

 

So you're saying it wasn't Stable?

[Coyotedancer]

I've been playing around with Bing's image generator for awhile now, and have had good fun with it... But I went in a somewhat different direction re: my City characters.

 

Back-in-the-day, I had a lot of winged "Bird Things"... and versions of most of them have ended up being pulled into our tabletop 13th Age game as NPCs. I've been using the image generator to make pics of those fantasy RPG re-imaginings.

 

This is the version of Grey Kestrel, my DB Scrapper, that you'd find wandering around the city of Horizon's version of the Chantry...

 

 

 

The only one of the current, Homecoming original characters that I've tried to do that "fantasy RPG covnersion" on is my favorite Thorn Wielder Broadsword Brute, Semnai. It did a pretty good job, I think, although it has NO CLUE... not one... what a proper Oranbegan Thorn Blade looks like. Poor Xhelethe. 🤣

 

Sem looks properly "Elmore-esque" for a stabby chick in leather not-exactly-armor. (Her City version spends most of her time in a very Emma Peel-inspired black leather cat-suit, so... yeah. I can totally imagine her wearing this outfit, too. ) I really wish it had picked a consistant style for the tats, though. City-Sem's work better there. 

 

 

[Techwright]

10 hours ago, sbloyd said:

I'm glad sharing our prompts has helped people, that was part of what I was hoping when I started a thread specifically for AI generative art separate from the other art thread.

 

I was experimenting with my "make coloring sheets" AI prompt, which is really just my usual prompt with "line art, black and white" added, when I got the idea to include "screen tone" in it for shading.  After generating a handful of coloring sheets I wanted to see what just adding "screen tone" to my normal prompt, and I got this result for my Grav/Kin Counterweight:

city of heroes, background constrained in a circle, wide angle, full body, superhero, style of artstation,, korean woman, yellow and charcoal grey uniform, infinity symbol insignia on chest, yellow sunglasses, charcoal grey leather bolero jacket, cupping a swirl of energy in her hands, pixie haircut, hovering in flight, screen tone, no cape

 

 

 

That is an incredible look!

[Trike]

3 hours ago, sbloyd said:

 

So you're saying it wasn't Stable?

 

Oh no, it was VERY stable. If by by "stable" you mean "turned off".