First of all, hi this is my first time making a post after being a long time lurker and occasional commenter.
I am no artist by any means so what I create is mediocre at best. I’m not terrible, but I am a far cry from being good. So I had the idea of making a sprite as best as I could and then using AI to generate an image based on my art. The game I am making is a long way from even a demo being available but I was wanting to gather people’s thoughts on whether or not this would be a good idea.
I also had the idea of releasing 2 versions that are a copy and paste of each other with the only difference being one would have purely my mediocre art and the other would have the enhanced ones that AI made from my art, and I would like to know people’s thoughts on that too.
As you can see there is a large difference and as I said I am not an artist but I feel like this method could bridge the gap I have when it comes to making games. That being my mid artistic ability.
Why handicap your artistic growth with AI if you are already aware of your personal shortcomings? Maybe it’s fine if you’re looking for a short-term solution, but I can guarantee you’d find it better to just suck it up and take the journey of self improvement when it comes to your art. With enough studies, I can promise you can put out stuff more suited to your interests than any AI module can, but that’s my two cents.
Also, people will respect your project more if they know you put in the work rather than to rely on a machine.
I think it’s a good use of tools and should be a good way to continue learning without bogging down your game progress. Since it’s based on your drawings, it should hopefully help with the inconsistent results in AI generated drawings while also not looking as generic as many of the models.
It’s quite a glow up, as well. I’d definitely play the version with AI assets.
Those images are very different in style. Do you at least feel comfortable editing the AI’s images and working with its style? If so, then you might be able to get a passable set of images out of it. But if you’re heavily reliant on only tweaking your prompts and input images, then you need to consider the possibility that at some point down the road, you might have to compromise on what kind of images you use, in a way you wouldn’t when drawing your own, simply because you can’t get exactly what you want out of the AI. Just keep that in mind when deciding how you want to invest your time.
Personally, as someone who has played a lot of games on this forum, both AI- and non-AI-based, I think it’d be best if you didn’t use AI. When I play a game, I like to look at the little details of what people have drawn, but when I know/realize it, AI all the details become meaningless because AI doesn’t do that, and it doesn’t have that same personal touch people have
As an artist, I again say no; it stunts your growth. It disincentivizes you from trying to improve your art, and this next part depends on the AI you’re using, but with its whole stealing art and not giving compensation or asking consent, I don’t think it’s a good thing for it to be associated with one’s own game.
I like it! Having to search for over a week for a model I like, I’d recommend trying out the maximum size images too, just to make sure everything still works at the more extreme end with your model choice and current AI-use skillset.
There are a lot of different reasons people give when they argue against stable diffusion image generation, but I think if you include both in your game you might be offering an olive branch to those who dislike it specifically because they think it gens up just mildly tweaked versions of other people’s images. It’s bound to be a time commitment though.
Even besides the standard anti-AI arguments, I would say no in this case. Pixel art is a very unique style, creating art out of restrictions (those being grid-based and a small canvas size). Using AI to ‘enhance’ that instead removes those restrictions to create a weird hybrid image. If you look closely at that AI image you’ll see all of the grid lines are actually wavy and irregular. It looks more like a standard image has been created with a vaguely pixel-y filter thrown over it.
And then there’s the canvas size, you’ve gone from 320x640 to 728x1448. Ignoring the strange scaling ratios, that again removes the restrictions that makes pixel art shine. Even with the limitations of your current art skills, I would much rather play a game with the sprites that you’ve made than whatever amalgamation that the AI makes. Plus, continually making sprites of different characters and weight stages will allow your own skills to improve naturally, and you’ll see the quality of what you’re making get better over time. Furthermore, like @someoneoutthere mentioned, when you’re the one drawing you get to decide what sizes, poses and expressions get shown.
Well, since you ask it on Weight Gaming forum, I guess your game should contain weight gain content. There is no issues with you using Neural Network to generate pictures, but the real question is if you are able to generate good pictures at all. First try to make multiple pictures of your character with different body shapes and post it here for us to judge. Generating pictures of default thin body shape on which 99% of all Neural Networks trained isn’t the same as generating progressive weight gain pictures.
If you are a mediocre artist that wants to be good, asking a machine to make your art good is deciding that you will never be good and I don’t think anyone should think so lowly of themselves. You’re also creating an expectation among the end user about what kind of images they can expect, such that any loss of access to such tools would block the entire workflow.
As others have pointed out, pixel art isn’t immediately comparable to other forms of art since it is a different technique, designed with limitations as well as because of them. I am perfectly happy to play a game with meh graphics so long as the graphics aren’t supposed to be the selling point and grew up in an era that was grappling with how important graphics were. Ultimately, graphics won, but that games that didn’t focus on graphics in that era still maintain cult followings today while early 3D titles aged far faster; stuff like rollercoaster tycoon only had pixel art, but they did more than enough with that pixel art to allow the rest of the game to shine.
So yeah, I have non-intrinsic reasons to prefer the pixel art but I also think that the issue here is less it and more mindset and self-perception. Do not let that stop you from creating.
I don’t know what you fed the AI other than the first picture itself, but it’s added a lot of detail that wasn’t in the original. The earrings, the handbag or satchel, the belt, the expression - none of that is in the base image. If the AI hallucinates that much detail in a single image, how on earth are you going to corral it into creating a consistent set of images over time? If you specifically told it to add these extra details, why even bother giving it a base image to work with in the first place?
I just wanted to thank everyone who responded. While i prefer the look of the AI image i do agree that I would be limiting my growth which leaves me torn if im honest. As on one hand doing everything personally means id eventually get better at the art side but on the other hand I know my skills lie elsewhere within game design and storytelling. So truthfully I still dont know which would be better, id only do this for character’s as im quite confident in my ability to make items and backgrounds.
Can you get the character to look similar in multiple images showing the transformation of weight gain, or will they be radically different in appearance between images? And how much time are you expecting to spend wrangling the generations and prompts to get the images to come out close enough to seem like the same person? If you’re spending time and energy on this, how much of a savings is actually present versus doing it yourself?
You own Image #1. You don’t own Image #2. Even though it is a “transform” AI generation, current legal findings hold that anything an AI puts out is public domain. You will have to put effort in to significantly hand-modify every image coming out of a genAI prompt if you want to maintain the ability to legally pursue people taking that content from you, and if you’re already doing that? Better to just make it by hand in the first place, so you know things like the anatomy ratios aren’t wonky in a way you didn’t notice at first.
Using genAI gates your usage behind the availability of the model. If they jack the price up, or the model gets discontinued? Now you can’t make anything like what you’ve been putting out anymore! That’s what genAI services are betting on - that they can make you dependent upon using them, such that you’ll have to pay whatever they decide to charge in the future.
Better to stylize in simplified manner, so what you’re putting out matches your skill level, and have it be an artistic choice. 8-bit it, it’s already pixels. Then you can use more and more detailed pixel-art as your skills improve, moving up to 16-bit and 32-bit styles. As long as it makes sense with the rest of the design choices, it will be generally accepted - nobody complains about the art in the Gnorp Apologue after all.
AI is terrible at recreating pixel art. Look at the right (viewer’s right) edge of the woman’s dress in the AI art, you can see that it’s kept the same staircase effect from the pixel art on what is supposed to be a full-resolution piece and it looks, if you’ll pardon my French, butt ugly. If you must use AI to enhance your own work, go for traditional full-resolution art rather than pixels.
Personally, if I had to choose between these two images, I’d choose the one that is 100% your work. The AI art is higher quality in some places, sure, but it’s not consistent. Her dress is especially horrible, as it seems to be trying to keep the larger pixel edges of the original work while rendering the rest in a higher resolution Stardew Valley-esque style. It makes the whole image feel like my internet cut out halfway through downloading it and it got corrupted.
You can always improve your skills. Don’t be discouraged just because there are other people out there who draw “better” than you, they probably had to practice for a long time, too. If you keep at it every day, or at least regularly, you will improve. Try comparing your art now to your art from a year ago. I guarantee that if you’ve been drawing roughly once a day for the entirety of that year, you will be able to see improvement.
AI cannot improve, not in the sense that a human can. The best you can do is give it more references to pull from, which can be considered theft depending on where you get the images, or make changes to its algorithm, which 99% of people will not have the knowledge to do.
I can’t stop you from using AI, but if you’re asking my opinion, it’s a bad idea. It prevents you from developing your own skills, and it can produce questionable results at times.
That is actually a good idea. I hadn’t considered scaling down and then working my way up, I have always just gone in piskel and done characters as 32x64 but scaling down to 8 bit may be the right call for my ability and scope of what I am wanting to achieve.
While I agree with you, art is not my passion or my strong suit. I drew as a kid and a teen but have hardly done much outside of the occasional pixel art during my adult life. I enjoy art but its something I rarely do so unless I do it more my ability isn’t really going to get much better hence why I had this idea. But I do agree with you that actually drawing myself and slowly building my skills would be the better option.
Good comments from the against-camp above. TheWell-Being has legit concerns - I’d recommend that if you do decide on this route you either run entirely locally (ie the software is revision and redaction proof, because you’ve got it running entirely on site) or get a 100% idea of what images you’ll need and doing the whole batch near simultaneously to minimize your exposure threshold.
I think having a functional-if-not-perfect understanding of software like GIMP (which is free) is a near necessity for using AI images instead of Art(ifacts of human creative energies). It could be used to help ‘smooth out’ the ‘pixel’ sizes through a pixelization transform targeting the areas where the SD model let you down a bit in that regard.
I’m resolutely neutral on AI-generated art vs. uh, manual art. I get to see it either way!
Like others said, right now you’ve got the AI transforming pixel art into sorta-pixel art, and I’m not a fan of its pixelization choices. Look at the jaggies around the bottom of her dress!
Of course, you can get better at manual or AI art with deliberate practice; up to you how to divide your effort.
If you want to focus on your pixel art, I’d suggest taking existing photos or drawings, mostly people, that aren’t symmetrical or “on-grid” and tracing them faithfully into pixel art.
If you want to focus on the AI part of the pipeline, first pick pixel art style and commit to it or something else (anime, low-poly, photo…) and commit to that. Then practice generating and editing a character through various actions and poses while maintaining consistency.
I’m… indifferent, to be honest. I won’t automatically hate something if it has AI. Like, the upcoming Zoopunk game, which is a sequel/prequel to the highly underrated Forged In Shadow Torch, uses AI generation for certain NPCs and features like building ships and weapons, and it isn’t gonna be a gamebreaker for me just as long as I can play an action game with a cool bunny or rhino. Not enough rhino games out there.
That being said, I’d still rather it not be in the game, especially for the game’s sake, since many have made up their mind, and there’s no particular reason I need AI for anything. Sure, I’m not a good artist, but to me, AI art generators are a novelty at best. They’re fun to dink around with, and I’ve generated a few things that I ended up hand drawing, like a cool fat skunk character I ended up generating got redrawn into a character I felt like drawing once, but beyond that, I do not believe it should be used for serious artwork.
I get it, though. Making art is fun, but it gets a little less fun the more steps you add. I know a lot of people who end up sticking to simple spritework because they love designing characters, but maybe not the multitude of processes required to truly make something great. That’s fine, but I know people will respect someone who goes the extra mile a lot more.
