• Most new users don't bother reading our rules. Here's the one that is ignored almost immediately upon signup: DO NOT ASK FOR FANEDIT LINKS PUBLICLY. First, read the FAQ. Seriously. What you want is there. You can also send a message to the editor. If that doesn't work THEN post in the Trade & Request forum. Anywhere else and it will be deleted and an infraction will be issued.
  • If this is your first time here please read our FAQ and Rules pages. They have some useful information that will get us all off on the right foot, especially our Own the Source rule. If you do not understand any of these rules send a private message to one of our staff for further details.
  • Please read our Rules & Guidelines

    Read BEFORE posting Trades & Request

Using AI for cover art

unfair

unfair

Well-known member
Faneditor
Messages
227
Reaction score
218
Trophy Points
53
I mentioned in the Apology thread that I was interested in the potential for AI for cover art, as a replacement for stealing fan art to use as covers. And I thought I'd share my quick exploration into the idea. I used Bing Image Creator, which is free and utilizes DALL-E to generate images (which you normally have to pay for if you go through the openai site)

I started with the prompt "movie poster for the film Wind River", and while the art is interesting, it isn't topical and the AI doesn't seem to know what a movie poster is, because there's no text on it.

https://imgur.com/IgSGZXU
https://imgur.com/QhdxzOn
https://imgur.com/XbpVMXv
https://imgur.com/bOw0sXl
https://imgur.com/VxuJBGS
https://imgur.com/DSTUlFL

After this I tried "cover art for the film Wind River" and this was a bit more successful, apparently it knows cover art is supposed to have text, but the images still aren't topical. Although it does seem like a similar font used in some Wind River material, which is interesting, not sure if that's a coincidence or not.

https://imgur.com/KuMnco6
https://imgur.com/MQEa7xJ
https://imgur.com/LeYKR1Q
https://imgur.com/qd69Ivs
 
Then I tried several variations of a more descriptive scene with the actors names, and it refused to give me an image, same with the character names, so you can't generate anything using people or characters that the AI thinks might be copyrighted, which is a problem if you're doing superhero films I suppose. But Wind River is a more generic premise so I tried something related to the plot, and it was much more successful - "cover art of a snowy mountain with pine trees, at night time with a full moon, with a small native american woman running in the center of the image":
https://imgur.com/u8LYAea
https://imgur.com/tcVC6YY
https://imgur.com/cVkkQwg
https://imgur.com/xnMRLKl
The first one in particular seems very on point and I may end up making it into a poster
 
All of these look pretty good to me. I never saw this movie, so don't know how on point they are, but they sure look good. I might check this service out for some of my projects and ideas.
 
My attempt at using it as cover art:
https://imgur.com/mPb4Lkm

Honestly the part that took the longest was just paging through fonts to find one that had a little less of a standard serif look to it.
 
Trying to get a cover for "The Gentlemen" without using the actors seems like a challenge, but let's give it a shot... first something generic to see what happens "cover art silhouette of a smoking gun":
https://imgur.com/EejUj4u
https://imgur.com/bnToh30


Seems like an interesting direction, let's try more specific... "cover art silhouette of a smoking gun and a special leafy cigarette" (can't use the word marijuana, it's blocked/restricted, have to hint to get the AI to generate the banned content)

https://imgur.com/mbSrtux
https://imgur.com/zWQw3cK
https://imgur.com/vZR3Oug
https://imgur.com/ioLLP9K
https://imgur.com/zriiRbr
https://imgur.com/fmFY1fb
https://imgur.com/Bj4b7sk

A couple of those definitely have potential as a poster - not perfectly topical but probably the closest we can get without being able to include the actors. If you wanted to you could probably photoshop the actors profiles with transparency into one or more of those without too much effort too. That's probably the key - generate the items you need and then add the actor images yourself.
 
One more example... I was trying to get a good cover for Battleship, and after half a dozen prompts and literally 60 images generated I gave up and did a quick and dirty mashup using these four generated images:

https://imgur.com/w4G02yU
https://imgur.com/sI9D0L8
https://imgur.com/HFTPwqB
https://imgur.com/rWWN0YT

And the end result was:
https://imgur.com/6OBFxyo

Some titles would be hard to do this way though... I was trying to make one for Resident Evil and half the time it would deny my request to generate images because it didn't like the content it was generating. I'm going to try Stable Diffusion when I have more time, maybe later this week, and if that one is more permissive about generating franchise related content I'll check back in with the differences.
 
Fascinating. Wouldn't resident evil be "SWAT team man/woman shoots zombie in a spooky old mansion".
 
TM2YC said:
Fascinating. Wouldn't resident evil be "SWAT team man/woman shoots zombie in a spooky old mansion".
Click to expand...
I was going for a scene with Alice and the Red Queen, plus zombies, I suspect it was either detecting violence or sexual content because I had one prompt that worked once and then refused to work another time, so it wasn't a blocked keyword it was something the AI itself was spitting out that was apparently problematic.
 
The Wind River covers have the standard style/look you get out of ChatGPT and Bing which is “adequate” but readily spotted. You can get professional results from Midjourney or Stable Diffusion XL and then add the text in yourself with photoshop or equivalent.
 
I've had a couple of stabs at AI posters, but I've never really been able to get what I want from them.
For example, there's this poster I made by just using basic shapes:
holeatthebottom-front-3-1698578499.jpg
When I tried to make this idea with AI generators using various prompts, I got a lot of posters that look really cool, but didn't really get what I was going for:
poster.jpg
poster2.jpg
poster3.jpg
poster4.jpg

​
Also, 'If Hole Clove' Starring the love-child of Aiden Gillen and Tim Allen:
If-Hole-Clove.jpg
 
I've just completed 30 pages comic book with AI and doing some cover arts should be way simpler than that.
I guess I will try it sooner or later.
 
WilliamRedRobin said:
I've had a couple of stabs at AI posters, but I've never really been able to get what I want from them.
For example, there's this poster I made by just using basic shapes:
holeatthebottom-front-3-1698578499.jpg
Click to expand...

Seeing that example I figured it was probably achievable, so I ran a couple tests. Obviously as MightyAttackTribble said you're going to end up photoshopping it slightly to create your final product. First I just started with "night small house silhouette with lighted windows and small moon in sky" to see what it would give me:
https://imgur.com/84wO35b
https://imgur.com/qqgmmKa
https://imgur.com/S98BQ2j
https://imgur.com/NlErlAY

Then I adjusted to "trick" it a little bit into giving me a smaller house by adding something larger to the image with the prompt "night small house silhouette with lighted windows on top of tall mountain and small moon in sky" and that got me pretty close to where I wanted to be:
https://imgur.com/85k48YF
https://imgur.com/CJje2EC
https://imgur.com/uHe8P1t
https://imgur.com/Q6Lp0iR

Then a quick and dirty photoshop with the most promising one gives me this:
https://imgur.com/y8DEMgB


Edit: And if you didn't want a flat color for the earth you could probably prompt it for "a cutaway of a mountain showing the dirt, rocks, and fossils" or something like that, and then use the results from that prompt as texture for your underground section, maybe even use a few different results. Just crop them and adjust color appropriately and add into the underground area.
 
Last edited:
unfair said:
I was interested in the potential for AI for cover art, as a replacement for stealing fan art to use as covers.
Click to expand...
So, instead of stealing one persons work, you'll let ai steal several at a time?
Not saying I'm against it, but the morality (if that's what you're going for) seems grey.
 
Ray Danger said:
So, instead of stealing one persons work, you'll let ai steal several at a time?
Not saying I'm against it, but the morality (if that's what you're going for) seems grey.
Click to expand...

I know some artists are against it, and it's a far from settled issue, but as far as I can tell you have to work pretty hard to get generative AI to spit out the source material it was trained on. I see it as a tool in the same way that being inspired by someone's art would be.

It's never going to be as good as a real artist, but it can spit out lots of options very fast which is useful for a non-artist like myself who just wants a base with a certain concept to work off of - especially since I may have a general thematic idea but not a plan for the actual style so it's fun to see a variety of options that further hone that into a concrete concept even if what the AI gives needs to be tweaked a little or a lot to get there.
 
unfair said:
Seeing that example I figured it was probably achievable, so I ran a couple tests. Obviously as MightyAttackTribble said you're going to end up photoshopping it slightly to create your final product. First I just started with "night small house silhouette with lighted windows and small moon in sky" to see what it would give me:
https://imgur.com/84wO35b
https://imgur.com/qqgmmKa
https://imgur.com/S98BQ2j
https://imgur.com/NlErlAY

Then I adjusted to "trick" it a little bit into giving me a smaller house by adding something larger to the image with the prompt "night small house silhouette with lighted windows on top of tall mountain and small moon in sky" and that got me pretty close to where I wanted to be:
https://imgur.com/85k48YF
https://imgur.com/CJje2EC
https://imgur.com/uHe8P1t
https://imgur.com/Q6Lp0iR

Then a quick and dirty photoshop with the most promising one gives me this:
https://imgur.com/y8DEMgB


Edit: And if you didn't want a flat color for the earth you could probably prompt it for "a cutaway of a mountain showing the dirt, rocks, and fossils" or something like that, and then use the results from that prompt as texture for your underground section, maybe even use a few different results. Just crop them and adjust color appropriately and add into the underground area.
Click to expand...
This is very cool! Which AI did you use for it?
 
You must log in or register to reply here.
Back
Top Bottom