What do you think about AI generated characters?
- Thread starter DreamCity
- Start date
- Jun 29, 2017
- 106
- 124
At one point, Ai characters look too perfect, giving the feeling of being unreal. That said, when most CG is from Ai, then there is a high chance, that their styles will resemble each other. I think these two are the biggest traps an Ai-artist can fall into: Creating images being too perfect causing unreal feelings, and to resemble other styles, when many use the same model and similar prompts.
To overcome these two traps, one must invest time into learning Ai and especially the prompts. This is not an easy thing. To do this, the very first step is to avoid copying prompts from others and/or to stop using prompt-helper mods/plugins. There will probably two types of game devs: the ones who will see the endless possibilities in Ai - and the ones who will use it as an easy shortcut to create game CG.
In the end, how much players will like Ai-generated CG in games, will depend on the game dev and how they intended to treat Ai-tools.
To overcome these two traps, one must invest time into learning Ai and especially the prompts. This is not an easy thing. To do this, the very first step is to avoid copying prompts from others and/or to stop using prompt-helper mods/plugins. There will probably two types of game devs: the ones who will see the endless possibilities in Ai - and the ones who will use it as an easy shortcut to create game CG.
In the end, how much players will like Ai-generated CG in games, will depend on the game dev and how they intended to treat Ai-tools.
Last edited:
I'm currently working on a project using AI generated images and some post-editing, so I can give some first-hand input:
PROS:
- Excellent quality/time ratio: with AI I can produce a good picture in a day. It would take me 2-3 weeks to draw the same concept (and with much lower quality) by myself.
- Errors can lead to new ideas: sometimes you'll have a concept in mind, and the AI will fuck up and add some weird element to the picture. While most of the time you'll discard it and edit the prompt to avoid getting it again, there's times when the weird element will start up the gears in your head and lead to a more interesting design.
It's a bit like playing the telephone game: you'll tell the AI something, and sometimes they'll reproduce some other, often silly, often interesting result.
- Focus on the game: By letting the AI do the graphics, you can spend time working on an actual game, and hopefully produce something with real engaging gameplay, instead of what many "games" are: a collection of pictures locked behind a bunch of clicks and filler text.
CONS:
- Money: If you want the best result, you need to train your own models, which requires rather expensive hardware.
The same goes for high resolutions: I can only create small pictures because my PC can't handle any more than that.
In the end, it's much cheaper paying an artist than building an AI-capable PC if you don't already own one.
- Prompts: After all, an AI isn't a human, and it won't understand a prompt like a human would: you can emphasize "black hair" in a prompt, and the AI may interpret it as adding pubic hair, or just making all clothes black, which results in having to add extra prompts, each with the possibility of more errors.
- Noise: Getting a good picture takes me an average of 4 hours (some may take under an hour, and some others up to 10). While it's not a particularly strong noise, that's 4 hours in average that I'll spend listening to the video card's fans spinning at full speed.
------------------------------------------------------
I expect AI to take over amateur lewd games within a decade: it's simply too convenient - AI art allows anybody with a good story to make their own novel, or any game if they take the time to learn some coding.
I would've never starting on my game without AI assistance, simply because I can't afford to be throwing money at art for a game that will likely earn me the equivalent of two coffees a month.
I'm sure that as video cards become more affordable, more and more people will start using AI for their graphics, and of course, AI will continue improving, reducing the necessary time to get a good result out of it.
And this is just the start. The future will not only offer AI cgs, but AI animations, AI 3D, and even AI video way down the line: no more spending a week on a 3D model only to end up with an uncanny valley monstrosity; just throw the fattest card at an AI and give it the script of your scene.
PROS:
- Excellent quality/time ratio: with AI I can produce a good picture in a day. It would take me 2-3 weeks to draw the same concept (and with much lower quality) by myself.
- Errors can lead to new ideas: sometimes you'll have a concept in mind, and the AI will fuck up and add some weird element to the picture. While most of the time you'll discard it and edit the prompt to avoid getting it again, there's times when the weird element will start up the gears in your head and lead to a more interesting design.
It's a bit like playing the telephone game: you'll tell the AI something, and sometimes they'll reproduce some other, often silly, often interesting result.
- Focus on the game: By letting the AI do the graphics, you can spend time working on an actual game, and hopefully produce something with real engaging gameplay, instead of what many "games" are: a collection of pictures locked behind a bunch of clicks and filler text.
CONS:
- Money: If you want the best result, you need to train your own models, which requires rather expensive hardware.
The same goes for high resolutions: I can only create small pictures because my PC can't handle any more than that.
In the end, it's much cheaper paying an artist than building an AI-capable PC if you don't already own one.
- Prompts: After all, an AI isn't a human, and it won't understand a prompt like a human would: you can emphasize "black hair" in a prompt, and the AI may interpret it as adding pubic hair, or just making all clothes black, which results in having to add extra prompts, each with the possibility of more errors.
- Noise: Getting a good picture takes me an average of 4 hours (some may take under an hour, and some others up to 10). While it's not a particularly strong noise, that's 4 hours in average that I'll spend listening to the video card's fans spinning at full speed.
------------------------------------------------------
I expect AI to take over amateur lewd games within a decade: it's simply too convenient - AI art allows anybody with a good story to make their own novel, or any game if they take the time to learn some coding.
I would've never starting on my game without AI assistance, simply because I can't afford to be throwing money at art for a game that will likely earn me the equivalent of two coffees a month.
I'm sure that as video cards become more affordable, more and more people will start using AI for their graphics, and of course, AI will continue improving, reducing the necessary time to get a good result out of it.
And this is just the start. The future will not only offer AI cgs, but AI animations, AI 3D, and even AI video way down the line: no more spending a week on a 3D model only to end up with an uncanny valley monstrosity; just throw the fattest card at an AI and give it the script of your scene.