Maybe it's because I don't know the workflow to do game character design, but it looks like the only AI used in the character creation here was the initial character design? Like, before in a normal process you'd get a human artist to draw you a character on that "model" sheet, giving you a front/side/back view of the character, and then take that 2d drawing and hand it off to a 3d modeler who would turn it into a 3D mesh and start taking the 2d drawings and turn them into textures, and then rigging them for animation controls.
It seems like all of that 2d->3d conversion is still being done by a human in this workflow, and so the "only" part that's being done is the 2d drawings. Now, my drawing skills would put me in the lower half of an average 4th grade class, so it's more than I could do, but it still seems like we're quite a ways from going from a couple of prompts to get a 2d design sheet, then a mesh and textures that match that 2d design, and then a rigged mesh, and then motion data to attach to that rigged mesh to give some basic animations.
So, while this is super-cool, it still feels like we're a long way from saying "I want a plumber with a thick mustache wearing blue overalls, a red shirt, and red cap. And now I want him to be able to jump and squat coming down." and start putting that into my game asset library.
There are ways to go from this art directly to 3D models and I’m sure tons of people will post links to them.
But I want to point out quality and artistic knowledge.
The poster is clearly a talented artist. They knew how to direct the images to what they wanted and likely rejected many iterations before they got to this point.
They knew how to then take that image and translate it to 3D , and make it come together. They knew how to make a topology flow that makes sense for their prototype etc…
I think a lot of people see “good output” and forget about the process and knowledge+experience required to make something good.
I take photography as an example. Every one with the means likely has a camera today. It still takes a brilliant eye to take an amazing photo.
You could take two brilliant photographers and someone without expertise to a gorgeous location. The two experienced photographers will have a higher likelihood to come out with amazing but very different images. The layperson could maybe take a good photo but the chances are much lower because they likely don’t have the experience to think about composition or story telling.
I liken these advancements in AI to that. Yeah, some people will put out pretty images, but there’s so much more that goes into making something great out of it that people underestimate
There is a lot of content in games. The AAA games can have single artists work on one character for the whole game. The author even says it saved him 5 days, and look how small the scene is he's talking about! AI has the ability to scale up the production quality of games, whilst taking half the time or less.
This kind of tech enables an indie to build a prototype with AI generated content, where the art direction is clear, then get funded in order to hire artists for the touch up. Kickstarters are often very visual, so this kind of stuff is required up front.
It seems like all of that 2d->3d conversion is still being done by a human in this workflow, and so the "only" part that's being done is the 2d drawings. Now, my drawing skills would put me in the lower half of an average 4th grade class, so it's more than I could do, but it still seems like we're quite a ways from going from a couple of prompts to get a 2d design sheet, then a mesh and textures that match that 2d design, and then a rigged mesh, and then motion data to attach to that rigged mesh to give some basic animations.
So, while this is super-cool, it still feels like we're a long way from saying "I want a plumber with a thick mustache wearing blue overalls, a red shirt, and red cap. And now I want him to be able to jump and squat coming down." and start putting that into my game asset library.