Will AI Destroy Animation As We Know It?
I’ve recently been bombarded with a lot of doomsday talk about AI’s disruption of the animation industry, often in articles with clickbait-y titles like this one. My old day job was all about disruption and technology curves, and I now spend every day knee-deep in the animation trenches, so I do actually feel qualified to talk about this. It’s a subjective and emotionally charged topic — with good reason — and I usually try to steer clear of both trendy controversy and unsolicited pontification. But I’ve had a lot of people ask for my take, and the answer is complicated — so against my better judgment, let’s talk generative AI and animation.
(Feel free, by the way, to skip to the end for the traditional 3 big takeaways one would expect from an article like this.)
I’m an AI optimist, to be clear. As a solo animated filmmaker, the idea of an AI assisting with rote tasks — but also being a tireless, always-on creative partner, generating endless options and suggestions, and massively speeding up my workflow — is wildly attractive. Right now this is far from a reality, but as a storyteller with lots of ideas, I would love a turnkey creative animation studio robot that executes on those ideas at a high level of quality and low cost. But mine is just one perspective.
There are other viewpoints involved, quite different from solo indies like me: artists at large studios producing bigger budget professional content, the executives that run those studios and the investors they answer to, and the general public that consumes all the end products. Over the last year or so, I’ve ended up in discussions with people from all these groups and the core arguments tend to follow a set of similar patterns. They’re categorized below (broadly) into ‘For’ or ‘Against’ the long-term impact of AI on the animation industry.
Lets examine them one by one:
Against: “They stole my artwork.”
This is a legitimate objection by artists whose work is now a part of The Matrix against their will. The original sin of unattributed, uncredited, scraped training data used without consent — most image generation LLMs, however much they tout their ‘responsibly sourced’ credentials, are guilty of some version of this. The question is how to fix it, and more importantly, if anyone cares enough to do anything about it. Unfortunately, without any legal or economic imperative, that seems unlikely to happen. Regulation and copyright law are woefully unequipped to deal with the complexities of this problem. Naturally the blockchain wants to fix it, but with web3 funding drying up and going to AI instead, I’m not optimistic about that either.
For: “But it’s just another tool.”
This is an oversimplification, but there is truth to it. Various kinds of Gen AI are destined to be embedded in existing workflows, eventually becoming “just another tool” that artists use to realize their vision. Where the analogy breaks down for me is that tools tend to do what you tell them to, but they don’t usually give you new ideas — that has historically been the domain of a human. I think it’s more helpful to think of Gen AI as a creative partner, with whom you can spar to develop interesting new concepts. A veteran animator friend describes it as “playing Director to the AI.” It’s good at coming up with options— you have to decide if they’re interesting. You are still in control, and the concepts that are generated usually need a lot of correction / extension (paint-over, photo-bashing, compositing for now — and eventually retopology, re-rigging, re-timing and more) that uses a traditional skillset. It’s a bit like having a talented intern who doesn’t have much taste, but also never sleeps.
Against: “The results are not very good.”
This is fair, though in a very narrow sense. “Intention” is a really important word in animation. The way a character shakes their head, the shape of their eyes, the color of the sky, and the framing of the shot — everything should be an intentional choice in service of the story. The way most Gen AI works is pretty different from that, so I understand some of the dismissiveness I’ve heard from certain quarters. Static images from Midjourney or Stable Diffusion might look really good by themselves, but as part of a larger narrative or project it’s hard to achieve the kind of consistency or control required. But what if you intentionally combined bits and pieces of AI generated content into something more cohesive?
For: “Combined with traditional skillsets, it’s a huge force multiplier.”
Think augmentation, not replacement. Using things like Midjourney for texture generation, or character design, or environments, really does feel like living in the future. It’s a matter of time before this extends (well) to 3D geometry and full 3D environments / rigged characters with optimized topology. Gen AI works fast and generates a lot of options — none are perfect, and usually need various kinds of fixing, but it’s still a significant creative and throughput boost. If the new workflow enables faster iteration, that is a huge plus in a creative endeavor.
Against: “Image generation is pretty good, but text-to-video is garbage.”
Subjective, but largely true for now. Runway’s tools are the poster children for text-to-video / video Gen AI. But you can’t do consistent characters from shot to shot with Gen-1, and if using the still image input with Gen-2, you can’t do directed motion — what comes out is pretty random. For anyone who animates through “traditional” means, current text-to-video offerings feel impractical and uncontrollable with limited real application — sure it’s fast, but who cares when you can’t get consistent results that look the way you want. That’s fine for a ‘mood reel’ or a not-very-coherent trailer. All vibes, no story. But for someone who has never animated before it’s like discovering a superpower, because the results, while hard to control, are miles better than anything they could do otherwise. So “garbage” is pretty subjective — though by any professional standards, current text-to-video offerings have a long way to go.
For: “But it’s improving so fast.”
True, and quite unsettling. A year ago I wouldn’t have believed the things that are possible now. Classical disruption often starts with a measurably inferior product attacking the lower levels of a workflow or value chain. The speed at which Gen AI offerings are improving absolutely blows my mind, and is only going to accelerate. It’s easy to ridicule Midjourney’s seeming inability to understand how many fingers humans have, but think about where it was a year ago. Don’t judge these tools by what they can do now — consider instead how fast they are improving, and extrapolate.
Against: “It might look good, but it can’t tell a good story.”
True for now. By definition most ‘storytelling’ results that I’ve seen are derivative at best, and laughably bad at worst. ChatGPT occasionally does surprising things, but the initial wow factor goes away when you measure it by professional content standards. This is particularly relevant right now given the ongoing WGA and SAG strikes. There’s a comparison to be made with shooting high quality video — something that used to be very expensive and difficult, and is now much cheaper. Just because it’s now easier to shoot high quality video doesn’t mean that many more good films are made. The technical barrier may be easier to cross, but the artistic barrier — especially for complex storytelling and character concepts — still remains. For now.
For: “Humans make a lot of derivative, mediocre content too.”
True. Popular content is accessible — it’s not often cutting-edge creative. It takes creative leaps, but calculated ones, within boundaries. Which is a lot like how a lot of Gen AI stuff works. Just because it isn’t creating stuff from whole cloth, doesn’t mean it won’t be “good”. The idea that all content created by humans using traditional means is wildly original is an obvious fallacy. Plus large portions of the general audience don’t care about (and often isn’t even aware of) the final 20% of quality. If you’re looking to push the boundaries of the art, it’s fine to obsess over every tiny nearly-invisible detail, but if you’re just looking to feed the beast with content, 80% is just fine — especially from a studio executive’s perspective.
Against: “It’s going to take away a lot of people’s jobs.”
True. That’s disruption for you — in the most classic sense. In Clayton Christensen terms, Gen AI is taking root in simple applications at the bottom of a market and will relentlessly move up market, eventually displacing established competitors. Businesses and the executives that run them have a responsibility to their shareholders to generate better returns, and whether we like it or not, animation studios are businesses. If your job is rote, you’re on borrowed time anyway. If your job is creative, but only a very small portion of the audience can discern the difference between good and bad, you’re in some danger. If your job is creative, and most people can tell good from bad in what you do, you’re safe for longer. A broader skillset keeps you more deployable and valuable — the more narrow your skillset, the more in danger you are.
For: “It might take some jobs, but not my job.”
Ehhh, we’ll see. Anyone making predictions about AI trajectories (including me) is probably wrong. All the advances in the last 18 months have happened faster and in more surprising ways than most people expected. This type of industry disruption tends to evolve team structures and roles. Traditional animation skills that are not tied to a particular software or tech stack, and experienced eyes / taste for telling good from bad will always be valuable — you might just be using them in a different way.
And finally: “There will always be a market for doing things old-school.”
Mostly true. 3D did not kill 2D or stop-motion. Handmade goods can still command a premium. The imperfection inherent to handcrafted animation does give it a very specific, intangible appeal. Up to a point, this can be imitated by a computer — but beyond that point it stops being cost-effective. The reality to keep in mind is that while the overall pie will likely grow, the share that uses more traditional methods will inevitably shrink. 2D rules on television today, but is rare for feature films.
Big Takeaways:
Traditional animation skills (core principles, body mechanics, acting, posing, choreography, visual storytelling etc) are not going anywhere. The power and beauty of animation has never been about the tools being used to make it. The skills that make a great animator are fully platform agnostic — they are embedded in the eyes and brains of people that know how a character should fall, how a mouse’s ears should twitch when they are sad, what a powerful silhouette looks like. The industry will evolve and everyone working in it will have to as well. Smart, adaptable people will find new ways to apply existing skills in roles and team structures that did not exist before. Check out the Light & Magic docuseries for a fascinating depiction of how ILM Model Shop artists, including the legendary Phil Tippett, found new ways to apply their skills during the CGI transition. New tech and workflows will result in pushing the boundaries of the artform. Incremental throughput will improve but that alone won’t necessarily bring down overall costs or increase ‘quality’ (just look at what happened with 2D and 3D). Good taste and experience — knowing what is good or bad, and why — will always be valuable.
Gen AI is not just the latest flavor of the month — it’s a transformational moment for the industry, just like the advent of CGI. It will have ramifications at every level, affecting how all kinds of shows are produced — including those using very traditional workflows. Don’t stick your head in the sand — educate yourself, find ways to apply new tools, and stay ahead of the curve if you want to stay relevant. Your job may not be going away tomorrow, but on an accelerating treadmill, running at the same pace means falling behind.
This is not the death of the art of animation. Technology democratizing the ability to create higher quality content is not new. “It’s just another tool” oversimplifies the argument, but it is what people do with it that counts. The functional definition of an animator will change, as it has throughout history. Traditional classifications and categories will change too, as they have before. There will always be purists that don’t accept new art forms and new ways of working as legitimate, and that’s OK. If the end result is interesting (for how it looks and what it says, not how it was made) to an audience, and provokes thought, emotion, and discussion, then that is what matters.
Like I said, I’m an AI optimist and as a solo artist, I’m excited about our new robot underlings. As with every new technology advance that further democratizes content creation, we’ll have to live through a “dogs on skateboards” phase where there is a deluge of stuff made by people just because they suddenly can. Apparently in the case of Gen AI, the equivalent is Wes Anderson trailers. That’s OK too — eventually people will find clever ways to make things that are actually interesting — and I am excited to both watch and participate.
