Generative AI requires Skills and Hard Work

Usually, I'm a nice guy. I roll with the punches, if I can. But, I am so sick of people's attitude toward generative AI. I'm talking about using generative AI for game development -- especially for artwork. It seems that there is a misconception that using AI means that you aren't doing any work and/or your art skills are weak. Frankly, that's bullshit! What goes into my game? There are four types of assets that I use in my game: Purchased Assets. I have hundreds of purchased assets -- especially from Itch.io. I'm a bit of a hoarder. Some I buy because I want to use them in a game. Some I buy because they make me smile. Some I purchase because I want to learn from the artist -- I like their style and may want to mimic certain elements of it. (I've never used another artist's work to train an AI!) However, these assets are not left untouched, they just save me time. They still need edits and palette conversion to make them consistent with the rest of the game and I do these all manually and with PixelOver. My own assets. Yes, I draw my own assets the traditional way. Mostly, isometric tiles, UI and Icons but I have also done some characters. My tools are Affinity Designer and Affinity Photo. For animation, I will sometimes use Aseprite because Affinity doesn't have timelines. Technically speaking, some of the assets that I make are not pixel art because they start as vectors, 3D models or digital paintings and then I convert them to pixel art either by hand or using PixelOver. These types of assets require the least amount of post processing because they are made to meet my needs and using my game's colour palette. AI assisted assets. These types of assets either start with a sketch (sometimes detailed and sometimes crude) that I've made, along with several control-net assets that I create in Affinity in order to get what I want. I will go through many iterations of prompt fine tuning, control-net experimentation, touching up the sketch so that the AI "understands" what I want and hours of AI drawing and me redrawing until I finally find something that is close. Sometimes "close" means that I have several results that I kit bash together using my own skills to make the image coherent. Or sometimes it means taking the generated image and redrawing as much as 80% of it. Then there is the palette conversion to the game's palette because even an AI that takes a palette as a parameter doesn't generate exactly those colours -- it just gets close. So, PixelOver and some other scripts come into play. However, depending on how picky I want to be, the colours the script chooses are not always good and so I will convert some of the important areas to my game's palette by hand, first, and then allow the scripts to figure the rest out. Consistency in quality and style are often an issue for AI, in my experience. AI is getting better and there are ways to reproduce a style in AI through Loras, seeds and style specific models. Although I don't use a style specific model (other than pixel art style), I will use Loras on occasion to add clothing elements like armour or monsters which the generic model may not understand. But most of the time, my little sketches get the job done. However, AI is not perfect and lacks a certain something sometimes -- especially for smaller elements in pixel art. It can generate a lot of noise and things that need to be simplified or removed. Unless the artwork's resolution is high enough, I can just about guarantee that character's eyes and hands are messed up. Even the legs and arms might seem off by 1 pixel. So I redraw and clean these up manually. So lots of work and no small amount of skill is involved. Honestly, I work my ass off to do what I do. Pure AI Assets. This is what I think most people imagine that I'm doing. Pushing a button and a perfect image comes out in seconds to include in my game. This HARDLY EVER happens and most of the time it is only in large background images. Even most of these, I have to touch up or heavily modify for the effect I'm going for. These assets account for an incredibly small percentage of my game. So, I don't get you! Here is what I find hypocritical about certain people's perception of AI. They have no problem with me buying an Artist's product or commissioning their work. But they have a problem with me using an AI to assist me. Even though there is work and effort involved to use the AI and lots of post work, I get labelled as not having skill or I hear expressions like "Isn't this AI?" as if there is nothing of my work in it at all. I don't take 100% credit for what the AI does any more than I would someone else's work. But when someone looks at your work and only focuses on the fact that AI was used and discounts everything you put into it, I think that's a demeaning and shitty attitude. Someone recently wrote that "Anything you make without using AI is 100% better" right afte

May 3, 2025 - 17:11

Generative AI requires Skills and Hard Work

Usually, I'm a nice guy. I roll with the punches, if I can. But, I am so sick of people's attitude toward generative AI. I'm talking about using generative AI for game development -- especially for artwork. It seems that there is a misconception that using AI means that you aren't doing any work and/or your art skills are weak. Frankly, that's bullshit!

What goes into my game?

There are four types of assets that I use in my game:

Purchased Assets. I have hundreds of purchased assets -- especially from Itch.io. I'm a bit of a hoarder. Some I buy because I want to use them in a game. Some I buy because they make me smile. Some I purchase because I want to learn from the artist -- I like their style and may want to mimic certain elements of it. (I've never used another artist's work to train an AI!) However, these assets are not left untouched, they just save me time. They still need edits and palette conversion to make them consistent with the rest of the game and I do these all manually and with PixelOver.
My own assets. Yes, I draw my own assets the traditional way. Mostly, isometric tiles, UI and Icons but I have also done some characters. My tools are Affinity Designer and Affinity Photo. For animation, I will sometimes use Aseprite because Affinity doesn't have timelines. Technically speaking, some of the assets that I make are not pixel art because they start as vectors, 3D models or digital paintings and then I convert them to pixel art either by hand or using PixelOver. These types of assets require the least amount of post processing because they are made to meet my needs and using my game's colour palette.
AI assisted assets. These types of assets either start with a sketch (sometimes detailed and sometimes crude) that I've made, along with several control-net assets that I create in Affinity in order to get what I want. I will go through many iterations of prompt fine tuning, control-net experimentation, touching up the sketch so that the AI "understands" what I want and hours of AI drawing and me redrawing until I finally find something that is close. Sometimes "close" means that I have several results that I kit bash together using my own skills to make the image coherent. Or sometimes it means taking the generated image and redrawing as much as 80% of it.
Then there is the palette conversion to the game's palette because even an AI that takes a palette as a parameter doesn't generate exactly those colours -- it just gets close. So, PixelOver and some other scripts come into play. However, depending on how picky I want to be, the colours the script chooses are not always good and so I will convert some of the important areas to my game's palette by hand, first, and then allow the scripts to figure the rest out.
Consistency in quality and style are often an issue for AI, in my experience. AI is getting better and there are ways to reproduce a style in AI through Loras, seeds and style specific models. Although I don't use a style specific model (other than pixel art style), I will use Loras on occasion to add clothing elements like armour or monsters which the generic model may not understand. But most of the time, my little sketches get the job done.
However, AI is not perfect and lacks a certain something sometimes -- especially for smaller elements in pixel art. It can generate a lot of noise and things that need to be simplified or removed. Unless the artwork's resolution is high enough, I can just about guarantee that character's eyes and hands are messed up. Even the legs and arms might seem off by 1 pixel. So I redraw and clean these up manually.
So lots of work and no small amount of skill is involved. Honestly, I work my ass off to do what I do.
Pure AI Assets. This is what I think most people imagine that I'm doing. Pushing a button and a perfect image comes out in seconds to include in my game. This HARDLY EVER happens and most of the time it is only in large background images. Even most of these, I have to touch up or heavily modify for the effect I'm going for. These assets account for an incredibly small percentage of my game.

So, I don't get you!

Here is what I find hypocritical about certain people's perception of AI. They have no problem with me buying an Artist's product or commissioning their work. But they have a problem with me using an AI to assist me. Even though there is work and effort involved to use the AI and lots of post work, I get labelled as not having skill or I hear expressions like "Isn't this AI?" as if there is nothing of my work in it at all. I don't take 100% credit for what the AI does any more than I would someone else's work. But when someone looks at your work and only focuses on the fact that AI was used and discounts everything you put into it, I think that's a demeaning and shitty attitude.

Someone recently wrote that "Anything you make without using AI is 100% better" right after I showed my artwork in my game. Honestly, that's like saying, novels written on a typewriter are 100% better than those written in MS Word. I realize we can become lazy and rely too much on technology like Autocorrect and forget how to spell. And I believe that is what the commenter was trying to say. But does the author's use of a spellchecker take away from the craftsmanship of the story written? If it takes skill and hard work to make assets, but it's just a different set of skills compared to other artists, why is that less than someone who does it the traditional way? What do you want me to do next, crush up bugs to mix my RGB values?!

So in short, using generative AI is hard work that requires skills! And don't let anyone make you feel otherwise!