I have been following the recent mini-controversy around AI art and I’m getting a little anxious about what this could mean for the anime industry.
For those of you that haven’t gotten around to this particular piece of news, let me try to give you some context. First I think it’s important to separate AI art (art generated through artificial intelligence) from CG art. All AI art is CG but not all CG qualifies as AI. At least not to me. In order to understand my thoughts in this post, the distinction is important.
Unlike a lot of my contemporaries, I am not someone who dislikes CG in anime. I do not always enjoy the effect and I do think bad CG can really be annoying, but I also think bad art can be annoying. The CG in itself is not the issue. Personally, I just see it as yet another tool for artists to use. Maybe one that hasn’t yet been perfected but certainly one that offers a lot of potential and utility. Ultimately I cannot blame an artist who chooses to use it.
On the other hand, the way I define AI art in my head is when the computer replaces the artist instead of the tools the artist uses. That is the essence of the debate around AI art at the moment. If you can just feed a vague description into a computer and get a piece of “art” instantly or for free or almost free, why would you need an artist? And indeed, it seems as if artists are already losing jobs to AI-generated art. The question is not entirely black and white. I can easily see a world where AI and artists work together for instance. That’s certainly something that’s been beneficial in other fields. However, the idea is that a machine cannot (for now and arguably ever) replace the sensibilities, vision, and instinct of an artist. That all AI-generated art can only be variations and combinations of existing art. (Yes we can argue that all art is that, to begin with. We just won’t argue it here because I don’t have the space). And if we start replacing most artists with AI for convenience’s sake, we will end up losing something very important in the process. Some people call it the death of art itself or creativity.
Others feel that art is about the observer and not the artist. That if you get as much joy out of an AI creation as you do out of a human-created one, then there is no difference. Commercial art has been created by humans for decades if not centuries. It is deeply naïve to believe that every artist pours their soul into each commission and imbues every work with meaning. Others still could argue that under capitalism art is at least on some level motivated, or incentivized, by profit, making the creative aspect secondary at best.
And this is where I bow out. The conversation is truly fascinating in my opinion and says a lot about us and what we value as societies. I have a lot of thoughts myself and occasionally they are contradictory. After all, I have a Picrew avatar on this blog. I enjoy 8-bit art. However, I do agree that supporting artists and just encouraging humans to create art, in general, is important.
But I don’t have the knowledge or proper vocabulary to really go in-depth on this question. I also didn’t think this is why you guys come here, to this blog. We’re supposed to talk about anime! And so we will.
To me, anime is art. You can see when a show was crafted with love and enthusiasm. The industry is notorious for its poor working conditions and so anyone that sticks with it for a long time must, on some level, love what they do.
BUT, the industry is also famous for trying to cut costs, often at the expense of the artists that create both manga and anime. If AI art became sophisticated enough that you could simply upload a digital volume of the manga and it would pit out a fully coloured and animated version of it, I’m absolutely sure studios would jump on it. And that would be disastrous. Wouldn’t it? Would it? Here’s the BIG problem, I can imagine a universe where it would work.
I don’t think the idea of feeding an AI with source material and getting at least a first draft of an episode out is far-fetched. Think the technology is probably already available and will likely be perfected and made more competitive in the coming years if there’s enough of a market for it. It’s not crazy Sci Fi, we’re all in a simulation sort of talk. There are practical implications here.
And I can see how this could be incredibly beneficial. We know artists in anime work under completely unreasonable conditions so offloading a part of their work to machines could make those conditions a bit more bearable. The art would be based on what’s already in the manga as would the story, as such a lot of the creative elements are already in the source material.
If you still have a character designer that translates the important characters from manga to anime and you still have artists to create key visuals and go through the drafts to correct elements and add flair, you would retain some of the artistic sensibility of the visuals. Moreover, editors and directors would still be there to create storyboards and smooth out the final product. Basically doing the hard work of the adaptation. You could use the AI to do the procedural task of converting the story from a manga medium to an anime medium and start out with a much better base, thereby cutting the workload without sacrificing too much on the final product. It’s a win-win.
This could also allow studios to put out a lot more anime so it’s a win for us viewers. Not for me. My particular brand of special would make it so that I would want to watch everything and get completely overwhelmed. But reasonable folks would have more options and less-known manga would get a chance at adaptation. YAY!
Heck, in a few more years we might even see the advent of fan adaptations. Loyal readers of the manga with a little bit of technical savvy get together to create unofficial anime. What a brave new world that will be! I can smell the infringement suits from here.
But that’s a very optimistic view of the potential of AI art and AI animation. Expect studios will happily start cutting jobs they can automate even if the end product suffers for it. And I should say, there are some deep systematic problems in the industry that puts a huge amount of pressure on anime to be produced quickly and cheaply. It’s difficult to entirely blame any particular studio when they are doing what they can to simply stay in business.
I think that key animators and character designers will probably be the first to go. Studios will still need animators but I figure they will have general roles and intervene when the AI simply couldn’t create anything viable. They will probably also be assigned to multiple projects at once giving them little time to pour over specific series. So any visual flare not already present in the source will be lost.
For instance, in the fourth season of Haikyuu, there was an art change that many fans did not like (I did but that’s irrelevant). This art change was also accompanied by a bit of animation eccentricity. In certain action scenes, the characters started being animated with completely exaggerated rubber physics. It looked amazing. It wasn’t in the manga though so a procedurally generated adaptation would not have included this unless someone went in to manually add it. That’s not necessarily the artist though. It was probably a director that chose to do that and then a team of animators and editors that brought t to life. I’m just saying that an AI wouldn’t have.
I also think that storyboards will quickly become at least partly AI-generated. That’s going to have a huge impact on the adaptation itself. Adapting manga to anime is a lot more difficult than it may seem and simply going for the most faithful adaptation isn’t always the best idea. Just because the same things happen in the same order doesn’t mean it’s going to feel the same. I doubt a series director will ever be completely replaceable for that reason However the more control is taken away, the less a director is going to be able to do.
And the most insidious part is that in my opinion, if a larger part of creating anime is automated, then people working in the industry will simply be asked to work on more projects with tighter deadlines instead of using that extra time to make each anime better than it would have been otherwise.
The fact is, although anime is a very commercial product, it’s also an art. In fact, it’s an amalgamation of a ton of different arts. We are getting to the point where we can have machines simulate all those arts but ultimately, it’s still just reproduction.
I have favourite studios, favourite directors, and artists and writers. I even have some favourite composers and voice actors. (I haven’t mentioned it but I bet voice acting is also something that can be synthesized.) The fact that I have favourites is proof that individuals bring something to the work they produce. There are anime that I really love but that weren’t made by my favourites. There are times when artists I really like missed the mark for me. And this sort of adaptability and variability is really important to making good art, in my opinion.
That’s why this rapid development of Ai art scares me a little. There’s not much can do about it though. But for now, I’m going to keep supporting artists in every way that I can.
37 thoughts on “My Fears for AI Art in Anime”
Ai art is certainly fun to play around with.
But I’m not sure how I feel about people selling the art that the Ai produces. It just doesn’t seem right to me, sort of like, for example, forking OBS and selling it. Why would you buy it when you can get actual OBS software for free. Just an example of why it seems weird to me…
It’s really a wonderful tool if you use it well
I just came across this article on ANN and thought it would be of interest to this discussion. https://www.animenewsnetwork.com/interest/2023-02-02/wit-studio-produces-the-dog-and-the-boy-anime-short-with-ai-generated-backgrounds/.194426
I think you’ve put your finger on the nub of the issue: while technology is a tool for an artist, then that is fine, even if the results aren’t always “great art”; but once the tool becomes the end in itself, then something profoundly important is lost, and “art” merely becomes another “product” (not in the commercial sense, but in the purely mechanistic “output” sense). At that point, I think it ceases being art and merely becomes an “item”. The implications for human beings – and for human activities such as art – are complex and largely unknown; and unless we are very, very careful in their application (ie: unless we ensure they always remain a means to an end rather than an end in themselves) I think you have good grounds for holding the fears you do.
I hope I’m completely wrong
I’m not too fussed about AI’s direct implications on anime production. Certainly, AI will render certain tasks redundant. As with previous technologies, internal processes will be streamlined & restructured. And efficiency is a goal worth striving for… So long as it allows for an increase in both output & quality.
But what’s special about AI, is its pervasive applicability. Some of us have seen globalization disrupt industries. We’ve seen automation do the same, on a slightly larger scale. Well, AI operates in a wholly different ballpark. We haven’t even begun understanding just how dramatically our economic systems will be impacted.
What we do know, is that high-skill folks tend to do well in these kinds of systems. They can leverage their talents more easily. They can improve their products & services, they can increase their reach & sales. On the other hand, lower-skill workers become more & more redundant. Their jobs could be displaced, their wages will be suppressed. And again, this affects every single industry, in every single part of the world, all at the same time. The scale is absolutely incredible.
That’s why I find the socio-economic implications of AI deeply concerning. The gap between working time & wealth generation will deepen. The mismatch in power between capital & labor will widen. Technology-induced structural unemployment will rise. The cohesion of our societies, already stretched, will be put under immense strain. If we want to successfully manage this transition, we all need to start asking ourselves some serious questions.
How can we ensure the economy works for everyone if, while we are producing more than ever before, more & more of it is going to a small group at the top? How do we make sure that every citizen has a living income? How do we invest in early-childhood education, or encourage continuous learning for workers? Should we reexamine our social safety net, or readjust our taxation system? And what does this mean in terms of us supporting important but underpaid occupations like teachers, child-rearers, caregivers, paramedics, artists? How do we collectively redefine what we value, and better reward the skills that truly matter to us? (Chief among which, anime. Amirite?)
We’re going to need to redesign our social compact, over the next 10 or 20 years. Hopefully we’re up to the task.
Re-reading my comment, I don’t know if I’ve gotten off-track. Sorry if so. As for the concerns you raise around originality, adaptability & variability… That’s exactly why human guidance & input will be required, over the medium-term. But probably not forever. At some point, machine learning will enable AI to clear these hurdles. Eventually, it will become sophisticated enough to produce the kind of audio-visual stories humans can resonate with emotionally. With the right data & processes, it could perhaps even understand us better than we do ourselves, tailoring whatever it creates to each individual’s personal preferences, or momentary moods. (That’s a natural evolution of existing social media algorithms, which are still fairly rudimentary in many respects.)
I find the implications of that truly scary. Not the AI art itself. If it’s written well, emotionally resonant, thematically intriguing… I don’t really care whether it’s created by a human, a robot or a chimpanzee. I disagree with other commenters on that. What scares me is, that means AI would be able to read & reach into our psyches much more deeply than it does today. This raises a lot of important questions about privacy, governance & corporate ethics. And I don’t know that we’re prepared to tackle them.
We have opposite fears. I actually don’t care about robots becoming sentient… I am pretty ok with the idea that most of our emotions/creativity… is up to a mix of neurochemical and can be manifactured. I still think it,s special. But I’m afraid that we’ll get lazy and not advance machine learning for actual creativity or jump mutations. We’ll lose the truly weird or subtlelly brilliant. And we’ll lose the imperfect. I think those are important to art.
My fear isn’t about robot sentience, though. More about AI being so powerful, that governments & corporations would be able to easily cross some lines, that perhaps shouldn’t be crossed. But that’s a tangential discussion.
I see your point. My best guess is humans will be involved, until AI has what you mention down pat. The market will require it. Most of us like anime because it’s weird or imperfect. Or because it comes up with some truly ridiculous ideas, like yakuza maids or zombie idols. At the end of the day, you gotta satisfy your audience.
Do you have to satisfy your audience though! There are so many products that are objectively not satisfying or worse than they use to be but we just accept them for questions of monopoly, convenience, price, availability, advertising…
Anime wouldn’t have a monopoly on audio-visual storytelling, though. I think live-action is here to stay, for a good while still. That’s a medium with a lot of staying power, that I’d imagine AI couldn’t ever fully take over.
If anime gets boring, it’s easy for me to switch over. (In fact I already do it from time to time.) And I know I’m not the only one.
I’m not sure your representative of the entire audience. I have a feeling some people watch animation at least in part because it’s not live action. Heck I know tons of people that fell in love with animes they had never even watched just based on the visuals
You’re right, I don’t represent the entire audience. But that’s the thing, you don’t need the entire audience. So long as a good chunk of market share threatens to pull away, then that gives Capital something to seriously think about.
We could be talking about 30% of the audience. Or maybe 20% of the audience. Hell, 10% would be more than enough. In such cases, Capital will always put Management’s feet to the fire. And that’s normally when things start getting fixed.
To grow your numbers, there’s only so much cost-cutting you can put yourself through. At some point, you gotta give folks what they want. Or risk losing some of your revenue streams to competitors that do — both within your industry & outside it. And you don’t want that to happen.
So, either you work on improving your AI. Or, you have a few humans complement it. (Which wouldn’t even cost you that much, if you wanna be Machiavellian about it. If there’s one thing we’ll never run out of, it’s passionate artists hoping to make a mark. In the AI era, that’s even more of an employer’s market.)
No if they are saving 80% in costs then 10% of the audience is not going to be worth it. Especially when you consider that a generation later no one will even know he difference. Look at what happened to hand painted cels. Everyone says it’s better but that didn’t sway the market one bit. If the majority of the audience is willing to accept a product, there isn’t that uch incentive to invest in making it better.
Well, folks who preferred hand-painted cels in their audio-visual stories didn’t really have anywhere else to go. What was left for them to watch, beside older series? Also, I think we both agree that we’re talking about something much more fundamental here, than hand-painted cels. Colours with a grainier texture, versus creativity across every single aspect of production. The latter is bound to cause way more waves among the fandom.
Finally, the hybrid scenario where humans would complement an advanced, but incomplete AI could be one where you’d save 80% of today’s production costs, while maintaining your entire audience. (At least while machine learning gets the AI where it needs to be.) I assume you’d only have to retain a few key artists, to help guide the AI & cover some of its holes — across all aspects of production. That should be a fairly minor expense, and would probably be a bargain in the grand scheme of things. As opposed to hordes of animators painstakingly hand-painting every single frame in the entire show, just to do the colouring!
So, high-risk, and low-reward. And the opposite for the hand-painted cels. I don’t think the scale is comparable between both scenarios. But what do I know.
They could have switched to live action like you suggested. And I don’t think the difference is that great. Hand painting cells imbued depth, created unique images, literally every frame a painting. If I’m to believe fans of the technique it was a huge artistic loss, at least to them.
I 100% do not think they would retain their entire audience. I doubt they would retain their entire audience even if they didn’t change anything. I also don’t think most anime fans would care one way or another.
I don’t know why I’m more optimistic about this than you are. It feels weird. That’s not how the world is supposed to work!!
Anyhow, it’s been a good chat, thanks for your time. I gotta sleep. Wish you a nice afternoon, and catch you later.
Good point. Everything will be ok! Talk to you soon Ramon!
I mean, do we need an economy? We know we’re at an excess of resources…
Absolutely agreed that this is a problem. I plan to look into the recent lawsuits filed against AI image generators and into the arguments for and against the use of these processes in the US, but anime is undoubtedly going to be affected as well, and almost certainly not in a good way.
AI art is likely to be a treasure trove in Copyright infringement precedents.
Yes, voice acting can be synthesized. I read about three months ago, how a company synthesized the voice of Vader, since the actor is getting up in the years.
Music is also no longer out of reach for the computers, though I don’t think they can compete with the likes of Yuki Kajiura or Go Shiina yet.
I think the takeover of AI in animation is almost inevitable.We should probably worry because Elon Musk may be right.
It will start in kiddie shows and then move to the mainstream as the process becomes more advanced. .Just as ChatGPT will eventually take over writing. That’s because, even if the work is formulaic, the formulas express what most people are looking for. Ninety percent of everything is dreck. That doesn’t stop it from being profitable dreck. Anime is already almost locked into specific tropes and character designs that barely change from one program to the next.
I hope humanity’s end will be something like, “Humanity has Declined.” Only instead of “fairies,” it will be AI apps. The alternative could be “Colossus: The Forbin Project.” Or something unimaginably worse.
My feeling is ultimately if art falls into mass AI generation, it will become so stagnant and dull that people will eventually revolt against it and human-created “real art” will rise again. Maybe that’s optimistic, but I’ve seen how AI anime-styled character generators have already “created” a generic-looking style possibly because it’s blending a ton of styles into one. Doesn’t look bad at all, but you can tell these AI-generated characters from others if you’ve seen enough of them.
I hope it will be something like the City. I’ve always loved dogs
Yeah artists getting replace by AI is becoming more and more real everyday. I’m really not looking forward to the day anime starts being created by an AI. Art is a form of communication, once a computer program takes over for an artist, the conversation stops.
I read a pretty compelling article by a group of artists explaining the situation. Strange days ahead
There are folk tales about the Devil disguising himself as a human to try to tempt people or steal their souls, or whatever, but there’s always something that’s still off about his human form, like he’s still got his tail or goat legs. I kinda feel this how it goes with A.I. art, because, like, you’ll put in ‘Sailor Moon’ as a prompt, and it churns out a pretty decent picture of Sailor Moon, but she’s got six fingers or three twintails, or something.
And, yeah, the technology is pretty cool to play around with, but if I ever wanted a usable image, I’d still pay for a human artist who understands how many fingers a person is supposed to have.
I agree with you but in the opposite way. Like the machine will always create the perfect Sailor Moon but humans sometimes screw up and add an extra finger. And that sixth finger is what makes it special and artistic. If that makes sense…
With the rise of procedurally generated AI art. Created from a list of traits or inputs from a creator and left to the AI to figure out how to deliver on said instructions, fear of it going everywhere including anime is of course to be expected. And I expect it will be used in Anime shortly… As the source of humor.
I expect anime creators will have AI do art then make fun of it in their series. Yes. Maybe they’ll use it to create the next Kaiju for Ultraman to fight or imagine what an alien cloned life form would be. But I hardly think they’d pin their entire budget and series hopes on a gimmick that at best is hit or miss. It’ll be used, yes, but it won’t be the end all be all of the creative staff… Unless your a small indie developer without the staff to do it any other way.
It seems to me AI art will be the forte of the Indy anime scene or small upstart studios looking to create without spending loads on a full team.
I think most people underestimate how much animation is already procedurally generated. They did a pretty cool explanation of it in Shirobako years ago
I’ll just say I really hope what you described in this article never actually becomes possible because that would just totally destroy anime as we know it. Granted, anime production has changed dramatically from its origins to the present day, but the idea of people being replaced by AI is an entirely different matter all together. I agree that this is indeed very scary.
I might be completely wrong but I do think it will be possible. It might not get adopted.
I agree AI and GC can be used as a tool but should not replace the artist;
we need to keep the human aspect in Anime.
I hate how rushed production has become for Anime.
I love watching an Anime that I know took painstaking time and effort to make.
For example there are few studios left that still paint their backrounds. I believe Studio Ghibli and CoMix Wave Films are some of the few that still do.
I’m not sure who still does cell painting and shading. I should look it up