Thoughts on AI Art?

Maverk

New Member
Jun 1, 2022
3
2
Personally, I like being able to ask a computer to show me EXACTLY what I want to see until the program produces a slurry image.
But I also understand it takes work away from artists, most of whom are either not full timing art or struggling.
Personally I don't have the money to pay an artist to make art for me so AI is personally a nice alternative.

In general I would like to ask about anyone's thoughts on AI art, but also AI art in reference to expediting ones earliest games.

My art is poor, I didn't take classes and have always used pencil and paper, not digital. I think that most people would prefer the AI art over my art but also many people would prefer an actual artist and not AI. That's my take.
 
  • Like
Reactions: dreamerquest

asen-de

Newbie
Dec 23, 2023
20
12
Ai art can look very detailed but also quite generic. It does not trigger anything for me personally. Actual art on the other hand, can be, even if not so well made, way more attractive.

See games like Shrift for example. The art there is not very good but still, it works..and it works for me personally better then every Ai art. So I think Ai-Art will not make any competition for actual artists on almost every level.
 

peterppp

Member
Mar 5, 2020
458
864
the problem with ai art currently - other than freak hands and obvious errors like that - is that the texture is smooth without details and imperfections, like porcelain or something. until it looses that unappealing smooth finish, it's of no interest to me
 

witcher234

Newbie
Dec 20, 2020
70
22
the problem with ai art currently - other than freak hands and obvious errors like that - is that the texture is smooth without details and imperfections, like porcelain or something. until it looses that unappealing smooth finish, it's of no interest to me
not true if you want imperfections you should state that in prompt and it will do it
 

peterppp

Member
Mar 5, 2020
458
864
not true if you want imperfections you should state that in prompt and it will do it
it's the finished art that is to be judged and it's not good enough with some imperfections if the rest still has that same smooth finish. everything has to be there, and isn't

put a mole on a porcelain doll and it still looks like a porcelain doll
 

observer80

Newbie
Jul 8, 2017
44
64
A lot of garbage AI being produced but there is some decent to good content out there and the main reason I support AI generators is that it is the only way I can get content of certain Anime characters from the 80's-early 2000's.

Most artists show no love for the older anime shows and just focus on what is current. Maybe you get 1-3 images a year if you are lucky, but that's disappointing and I don't wanna wait another 20 years for someone to make something of a particular character.

I don't have $50=$200 to shell out on an image to some artist and that's assuming they go ahead and accept the commission. My only alternative to find artwork on characters nobody is drawing anyway is AI.
 
Last edited:

exer

Member
Game Developer
Aug 16, 2017
114
286
As an artist, I did a bit of research into AI imaging so that I could better understand it and how it works. I didn't want to jump on the hate train if it wasn't warranted or just parrot what other people are saying about it.

In my experience, it's always worse than a real artist. Even if you get something looking polished and anatomically correct, it still looks soulless, generic, and steps into the uncanny valley. On top of that, it's also just an annoying process that boils down to getting lucky. You type in what you want to see, then basically just reroll until it spits out something similar enough to what you wanted. You can tweak the prompts to try and nudge it one way or another, but it's ultimately up to luck. I generally got 1 ok image roughly every 50 or so. Which can take a while depending on things like your computer, what model your using, etc. It's also very difficult to get multiple pieces to look like the same character. I wouldn't say it's completely impossible, but it generally takes significantly more tries. Since you need to get lucky enough for it to be usable, but also have the exact same hair style, clothes, proportions, colors, etc. Even then, they aren't going to look exactly the same. They'll just look close enough. Also, when using AI there's a lot of repetition. AI can only really replicate what it has, so you're going to end up with a lot of images that look very similar, which honestly isn't what you want for a game.

Sometimes online generators have been tweaked to help get specific styles more consistently, but it's still up to luck what comes out. You could easily spend as much on one of those to get an okay image that you would just paying a real artist. I'm just assuming on that though, it could easily end up costing much more. I only tried a few, and only generated the max number of times for free (usually around 5), and only got something remotely usable once.

Ignoring all of that, using AI will alienate a lot of people. Some people are against it ethically, and for others they just don't like it because of the uncanny valley aspect of it. Either way, you're going to be severely limiting your audience if you go the AI route.

In regards to using it as reference, it's always better to use real life. Googling isn't really any different than using AI to generate an image. Either way, you're typing in what you want to see. But using a real reference will ensure that you get the right details. I know some artists who use AI for reference sometimes when they can't find a real one that matches what they want, but these are very experienced artists. They can do things like fill in the missing details and correct things easily and quickly. So unless you can do that, then using AI for reference is just going to make the end product worse.

If you're looking to make a game, I have a few possible recommendations on how to go about it. From hardest to easiest.

1) Use making a game to get better at making art. If you have an ongoing project, you'll be consistently drawing and having to learn how to make what you want, which will help you get better. Your artwork starting out will be bad, but you can always go back and redo it. A lot of games have done this, and plenty of people's first games have had bad artwork. Most people's first games are just straight up bad, but that's just how learning works. You're bad at something until you put in the time and work to learn, then you get better. So if you want to make a game, and want it to be 2D, I'd recommend going this route.

And for the record, this is what I'm doing with my current game, Waifu Wonderland. I'm not an inexperienced artist, but I'm not particularly good at anime. I'm way better at realism and western style artwork. That's generally what I make for my SFW accounts. But I've always wanted to get better at the anime style and also thought it would be more time efficient and attract more people than doing more realistic work. So I spent some time to develop an anime style, and have been using Waifu Wonderland to get better with it. Some of the art is better than others for sure, but I'm getting better and more consistent. Once I have a solid system and have gotten completely comfortable (and have the game out and updated once or twice), I'm planning on going back and redoing the rougher looking artwork.

2) And then for the other option, do what other people who don't draw do, use 3D. Learn to use DAZ if you want the absolute easiest path, or learn to use Blender if you want to have higher quality work and/or animations. There's plenty of free resources and models for both.

3) Use Virt-a-mate, or VAM. VAM is technically a VR game, but it can be used in desktop mode to make and animate 3D models. It's got a little bit of a learning curve, but I'd say it's a little bit easier than DAZ. You can basically just download models and poses and everything like with DAZ, but then you can animate with just physics. It is lower quality than DAZ though. It uses DAZ's genesis 2 models, while DAZ is up to Genesis 9. So there's a noticable difference in quality, but it's easier. I made a game with VAM and people liked the animations well enough. There's a dip in quality, but it's still better than stuff like honey select or koikatsu. Which I wouldn't recommend using for the record. To get anything other than just the canned animations is more work than just learning VAM or DAZ, but it still looks worse than both VAM or DAZ. And the canned animations have been used so much that people can spot them a mile away and will run far away from the game.

4) And lastly, the laziest way to go about having artwork for a game. Use real porn. Assuming you're just clipping a few seconds of videos and/or using stuff that's already available online for free, I personally don't see anything wrong with making this kind of game. It's hardest to monetize this kind of game because of Patreon restrictions, and there's a very small group that doesn't like it, but it's not impossible. It's also a great way to practice making a game. The programming aspect is a whole other skill that's needed, and not having to worry about making anything will help you learn that process much faster.

The unfortunate reality is that there is no quick and easy way to get quality work though. One way or another, you have to either put in the work to learn, or pay someone who has put in the work to do it for you.
 
  • Like
Reactions: Lord Executor