I’m sort of afraid to even bring this up in case Blizzard hasn’t thought of it yet, but for those unaware, a recent development in AI art generation is that you can now plug in existing stuff (characters, art styles) instead of getting random results.
(Here’s a video showcasing it, where they create a fully cohesive stroyboard too: https://youtu.be/W4Mcuh38wyM)
Now that OW2’s revenue will come massively from skins, one of the few silver linings in my eyes was that more great artists are gonna get work, valued work at that. Except with this new tech, imo there’s nothing stopping companies from just cranking out skin designs from the neural network - in fact they’ll have a huge financial interest in it. It still needs 3D and animation work, yes, but the concept stage can almost be replaced, and we might not even be able to tell the difference. Many things in design are already derivative.
Thought it’d be interesting to bring this up and perhaps hear other people’s thoughts on it. And hey, support the artists directly if you can
I think the reality is that artists will be stuck for longer on the same project, making their job much more boring.
Apparently mythic skins take a full year to make, and some poor git has to presumably keep iterating on the same theme until Blizzard deems the skin FOMO-inducing enough.
Also remember that more revenue through skins doesn’t mean artists will get paid more, it means perhaps there will be more hired or perhaps existing artists will be on the overworked side from now on.
As for AI generated work….it will almost always be horrendous. You will always be able to tell the difference, and ‘most things in design are derivative anyway’ is kinda missing the point.
Derivative or not, a human artist has the ability to make things in context, make a joke, make a contrast, etc. AI can mash stuff together from set parameters, but the outcome will always be superficial with no real thought behind it and nothing more than random.
The basic gist is that you can’t replace artists and their thought processes and get quality work at this point in time.
Because these skins will have so much income relying on them, they need to be more on point and ‘worth’ the money than ever - not procedurally generated.
I did simplify in my OP. Here’s what I mean by “the concept stage being almost replaced” (and this is a massive guesstimate):
One human and the AI generator might be able to do the work of 3-4 current concept artists.
The technology obviously still needs oversight, but I actually wouldn’t call its results “horrendous” (which is scary). If I was asked to bash out 10 variations of a set skin design, I’d probably use it at least as inspiration, but it’s actually good enough that with some adjustments I could probably submit the stuff as-is to my boss. I recommend you watch the video I linked if you haven’t yet.
Is it art you created if you didn’t actually work to create it (I dropped my paint bucket but it made a really cool splash, tie dye shirts)? What if you didn’t intend to create it (mindless doodling leading to a cool result, an accidental strum on an instrument)? Is it art you created if the inspiration was by chance - e.g, not yours, you saw something in the world? Is kitbashing a bunch of components you didn’t create into something new still your art?
To me the answer to all of it is: it doesn’t matter. To me it’s about being able to realize a creative vision. Everything else, the means, effort, inspiration etc, that’s just a preference. Some people hate digitally produced music with a passion, let them.
It’ll get scary when AI creates convincing impersonations of real humans (like a 100% realistic deep fake video), but until then,
I for one welcome our fledgling silicon based concept artists … for now …
I agree. If you’re not a jobbing artist yourself, it won’t particularly affect you (hell, you might even enjoy more, cooler stuff being produced). Although if Blizzard does indeed use an AI generator and let go of artists, you might stop and think about the price they ask for their cosmetics.
But yeah, “pics or it didn’t happen” is definitely gonna become a pretty archaic phrase soon.
I just don’t think that AI is good enough to emulate the creative process.
If your boss asked for a funny skin where Torb is Santa, sure AI could mash together a Santa and Torb but it wouldn’t have the critical thinking to create a smart design that was funny.
I could see the the benefit of perhaps having AI adding less important details to save time, like different hairstyles or whatever…but I still don’t like it.
I think it throws up a lot of deeper questions. What’s the point of art if it isn’t a human expressing themselves? Do we want art churned out by computers for the sole purpose of making money? Would that art still resonate with a human?
Personally I want humans making art, commercial concept art included. There’s something so depressing about AI randomly generating it for us - like cutting off yet another avenue for people to communicate their experience to others.
I hope you don’t mind me asking, but do you work in the creative field?
I think this isn’t always needed as much. They just need something that will sell. But it’s hard to judge at this point.
To make companies like Blizzard money. This is an industry, not an art gallery. Don’t tell me you haven’t seen a “myriad” of soulless, cynical creations get airtime over the years yourself. It’s great that you’re only interested in human-made art, but there’s plenty of buyers for the grey sludge (and the sludge actually improves year by year). I absolutely agree that it could end up being rather depressing in a number of ways.
Just a pet dream of mine, it’s only a hobby for me. (You can probably imagine now why I made the thread)
This cohesiveness is what Stable Diffusion can do a lot better now. You can plug in “Widowmaker” or “Reaper” with some reference pictures, and you can play around with them endlessly. You can plug in a work-in-progress new character, and get a load of variations and poses without having to manually draw them- well, maybe the hands.
(I kind of really shouldn’t be advertising this lol).
The stuff you made is pretty neat, btw - they don’t “jump out” to me as super-duper unique (I hope you don’t mind me saying) - but it’s definitely a fine proof of concept, and playing around with some parameters is easier than redrawing something, right?
Then again, aren’t there humans who we consider artists who essentially “just” reconfigure the thousands things they’ve seen already, sometimes creating something we consider “unique” purely because it’s a new combination?
No, I agree that the line is murky. It’s why it’s important to talk about. I for one think that losing the human element of art would be a travesty for culture.
Ya the job security aspect of it and the premium you’re expected to pay for an artist’s creation is really the only point of contention from me.
Royalties, copyrights etc is a (likely) necessary byproduct of the fact that we gotta work for a living, but your view can change a lot depending on if you actually think that kind of stuff is necessary in a world where an artist doesn’t have to sell art to feed their family or pursue their dreams.
More than just art is seeing that issue though. Truck drivers are most likely the next to go, and while robotic automation of more mundane tasks like food service or machine generated software is still a long ways away, it’ll come eventually. As it turns out, there are lots of things humans suck at, but we put up with it because we need to eat. Ai art gets good enough, maybe it ends up being one of those things we end up sucking at.
Utopian or dystopian as it sounds there’s gonna be a point in time where people really start having to consider “what is money actually worth and do we still need it”
Hopefully not in my lifetime. Previous attempts didn’t end well
AI art can be used as a tool, and I think people forget there is more value when it is used as a tool BY artists. They can make concept art, using a quick render, and alter it enough to make it their own. AI is still largely a randomizing event, it has a real hard time being consistent. If you type “Mickey Mouse in the style of Gil Elvgren” you will get a different version, a different Mickey every single time. But an artist would use that image generated and consistently and on model iterate on it in a more detailed way than AI will ever be able to do.
you can probably tell by my stance on the topic.