A Cold Handshake: Why Commission Animation from People, Not Machines

In 1970 the Japanese roboticist Masahiro Mori described a strange and familiar kind of eeriness. A prosthetic hand looks like a real hand. Then you shake it in greeting. The hand does not squeeze back, and it is cold. In an instant, warmth gives way to unease (Mori 2012 [1970]).

Mori called this the uncanny valley (Japanese bukimi no tani; on the Estonian term, õõvaorg, see Saar 2015). The more something resembles a human being, the more warmly we feel towards it – until, at some point, the resemblance becomes frightening. An almost human-like figure can be more unsettling than one that does not resemble a human at all. A toy robot seems sweet to us, but a prosthetic hand already feels eerie.

Mori also noticed something that bears directly on our field: movement amplifies the effect in both directions. The pleasant becomes more pleasant and the eerie more eerie. A moving prosthetic hand is eerier than a still one (Mori 2012 [1970]). And animation is the art of movement.

Animation means giving things a soul

The word “animation” comes from Latin. Its root is anima – soul. An animator does not simply make things move; they bring them to life. A character takes a breath before saying anything. A bouncing ball flattens for a moment because the animator wants to convey its weight and elasticity. A hand stops in mid-air because the person hesitates.

At this point the reader might expect us to talk about the machine’s technical errors. We are not going to count fingers, as if the ability to count were the highest human faculty. Those errors will disappear with the next models, and the argument will disappear along with them. The point lies elsewhere. In a machine-made video everything may move flawlessly, yet nothing is alive, because nobody has decided why it moves. A motion has no intention, a pause no reason, a glance no aim. The viewer may not be able to say what is missing, but they feel it. It is precisely this vague unease that is the most dangerous, because there is no arguing with it. It simply sticks to the brand. The hand may be perfect, but it remains cold.

There is a scene in a 2016 documentary in which developers show Hayao Miyazaki, co-founder of Studio Ghibli, an animation created by AI. The AI has been trained to move human figures as quickly as possible. It feels no pain and does not know that a head needs protecting. So the figures writhe grotesquely and use their heads like legs to get around. The presenter suggests that this kind of movement could be used in zombie games. Miyazaki recalls a friend of his with a disability, for whom even a high five is hard, and says that whoever made this does not know what pain is. He calls what he has seen an insult to life itself (Arakawa 2016; Ling 2016). His objection is not a technical one. The machine moves the figures efficiently, but it does not know what movement costs a living creature.

The taste of the average

Lev Manovich, one of the most influential theorists of new media, wrote that database and narrative are natural enemies (Manovich 2001: 225). A database presents the world as a list and does not put it in order. A narrative creates a chain of cause and effect.

Generative AI is a child of the database that has learned to imitate a storyteller’s voice. The writer Ted Chiang compared ChatGPT to a blurry JPEG of the whole web: the overall picture is recognisable, but the details have been lost in compression (Chiang 2023). In a later essay he proposed the generalisation that art is born of many choices. A ten-thousand-word story involves roughly ten thousand choices, whereas a hundred-word prompt to a machine involves roughly a hundred (Chiang 2024). The machine makes the rest of the choices. It chooses whatever is probable according to its training data.

Manovich is not opposed to AI. Together with Emanuele Arielli, he wrote a book on the aesthetics of AI between 2019 and 2024 (Manovich, Arielli 2024). That makes the question he raised as early as 2017 all the weightier: will automating aesthetic decisions reduce cultural diversity over time (Manovich 2017; see also Manovich 2018)? When everyone draws from the same well, everything starts to taste the same.

The philosopher Byung-Chul Han writes about the aesthetics of the smooth: a smooth surface pleases, but it neither wounds nor offers any resistance (Han 2015). Many AI-generated images seem like the perfect fruit of this aesthetic. Glossy, soft, lit like a toothpaste advert. And forgotten before the screen goes dark.

The medium is the message

Marshall McLuhan’s famous phrase “the medium is the message” means that the effect of the medium itself is often more significant than the content it carries (McLuhan 1964). For a brand, this means something simple. A handwritten card says, before it is even read: you matter to us. If a viewer perceives a video as a cheap, automated substitute, they may hear this message in it: you weren’t worth any real effort to us.

This can be seen in real life, too. For Christmas 2024, Coca-Cola had a new version of its 1995 Christmas advert “Holidays Are Coming” made with AI. Interestingly, viewers in ad testing rated it highly. Public debate, however, went the other way: according to critics, AI had sucked the warmth and joy out of a Christmas classic (Kiefer 2024). For a brand whose global marketing slogan since 2021 has been “Real Magic” (The Coca-Cola Company 2021), this was a painful irony.

A year later, in December 2025, McDonald’s Netherlands released a 45-second Christmas advert made with AI. On social media it was called creepy and inauthentic. Three days later the company took the advert down, calling the episode an important lesson (Aliieva 2025; Cress 2025).

Measurements back up these experiences, too. In December 2024 the market research company NielsenIQ reported on a study in which more than 2,000 participants watched AI-generated adverts. The viewers themselves recognised most of them as AI-made and found them more annoying, more boring and more confusing than ordinary adverts. Measuring the brain activity of around 150 participants showed that even AI adverts regarded as high quality produced weaker memory activation. The researchers warned of a negative halo effect: a bad impression of an advert may carry over to the brand (NielsenIQ 2024). It is worth noting that this is a company press release, not a peer-reviewed scientific article.

Aura and responsibility

In the mid-1930s Walter Benjamin wrote about the aura of the work of art. Aura is a work’s unique presence: the work exists here and now, in one particular time and place. In his view, the aura withers in the age of technological reproduction (Benjamin 2008 [1935]). AI goes a step further: it reproduces an original that never existed. If an image has been created without a human being’s direct creative presence, the question arises of what becomes of the aura Benjamin described.

A brand, too, is a promise that someone is there. When a government agency explains a new service or a hospital instructs a patient, every frame is a claim. Someone has to check whether the arrow points the right way. The machine is not accountable. It does not even know it has claimed anything.

There is also the question of ownership. According to a 2025 report by the US Copyright Office, AI-generated output can be protected by copyright only if a human has, to a sufficient degree, decided for themselves what it expresses and how. Merely giving instructions (prompts) is not enough (U.S. Copyright Office 2025). Under the Estonian Copyright Act, the author of a work is the natural person or persons who created it (AutÕS § 28(2)). With a mascot generated entirely by AI, it may be more difficult to secure copyright protection and exclusive rights to its image.

What a studio really sells

People think an animation studio sells drawing. In fact, we sell decisions. What to leave out. Which gesture a character uses to introduce themselves. How long to pause before an important line. What colour worry is, and what colour the solution is. When there is a place for a joke and when there isn’t. Every frame is the answer to a question someone has asked themselves beforehand.

Tools change, as they always have. But it has never been the pencil that makes a picture a good picture.

Our motto is “Show, don’t tell” (Lepik 2025). Showing presupposes that someone has understood first. A machine can create an image. But a human being has to answer for what the image means.

Every time a customer watches your video, they are shaking your hand. Make sure that hand is warm.

Frequently asked questions

How do you make an animation or video with AI?

There are lots of tools and they change quickly, so we won’t give step-by-step instructions here. Before you start, though, it is worth answering three questions: what the video needs to say, who is responsible for the content being correct, and who owns the finished video. If the video speaks on behalf of a brand or an organisation, these questions matter more than the choice of tool.

Why might an AI-made advert harm a brand’s reputation?

Viewers often recognise a machine-made advert. In NielsenIQ’s 2024 study, viewers found AI-made adverts more annoying, more boring and more confusing than ordinary ones. The researchers warned that a bad impression may carry over to the brand. Coca-Cola’s 2024 AI-made Christmas advert drew a lot of criticism, and in December 2025 McDonald’s Netherlands took down its AI-made Christmas advert three days after releasing it.

Who owns an AI-generated video, image or mascot?

It is not always clear. According to a 2025 report by the US Copyright Office, AI-generated output can be protected by copyright only if a human has, to a sufficient degree, decided for themselves what it expresses and how. Merely giving instructions is not enough. Under the Estonian Copyright Act, the author of a work is a natural person. With a mascot generated entirely by AI, it may therefore be more difficult to secure copyright protection and exclusive rights to its image. If this matters to you, ask a lawyer for advice.

What is the uncanny valley and why is it relevant to animation?

The uncanny valley is a phenomenon described by the roboticist Masahiro Mori in 1970: a figure that looks almost, but not quite, human makes us feel uneasy. According to Mori, movement amplifies this feeling. Animation is the art of movement, which is why the uncanny valley is particularly relevant to it.

What does an animation studio offer that AI does not?

Decisions and responsibility. Together with the client, the studio decides what to leave out, which gesture a character uses to introduce themselves, how long to pause before an important line and when there is a place for a joke. Every frame is a deliberate choice, and someone is responsible for making sure the video’s content is correct. This is especially important for government agencies and the healthcare sector, whose videos need to be accurate.

References

Aliieva, E. (2025, December 11). Not lovin' it: McDonald's pulls AI-generated Christmas ad after social media backlash. NBC News. https://www.nbcnews.com/world/europe/mcdonalds-ai-generated-christmas-advert-social-media-backlash-rcna248590

Arakawa, K. (Director). (2016). Owaranai hito: Miyazaki Hayao [Never-Ending Man: Hayao Miyazaki] [Documentary film]. NHK.

AutÕS = Autoriõiguse seadus [Copyright Act of Estonia]. (1992). RT I, 12.07.2025, 9 (consolidated text in force from 1 October 2025). https://www.riigiteataja.ee/akt/112072025009

Benjamin, W. (2008). The work of art in the age of its technological reproducibility, and other writings on media (M. W. Jennings, B. Doherty & T. Y. Levin, Eds.). Belknap Press of Harvard University Press. (Original versions of the essay 1935–1939.)

Chiang, T. (2023, February 9). ChatGPT is a blurry JPEG of the Web. The New Yorker. https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

Chiang, T. (2024, August 31). Why A.I. isn't going to make art. The New Yorker. https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art

The Coca-Cola Company. (2021, September 29). The Coca-Cola Company unveils new global brand platform for Coca-Cola trademark [Press release]. https://investors.coca-colacompany.com/news-events/press-releases/detail/1036/the-coca-cola-company-unveils-new-global-brand-platform-for-coca-cola-trademark

Cress, L. (2025, December 10). McDonald's Netherlands pulls AI Christmas ad after backlash. BBC News. https://www.bbc.co.uk/news/articles/czdgrnvp082o

Han, B.-C. (2015). Die Errettung des Schönen. S. Fischer.

Kiefer, B. (2024, December 17). How Coca-Cola's AI holiday ad went from praise to rage. Adweek. https://www.adweek.com/creativity/how-coca-colas-ai-holiday-ad-went-from-praise-to-rage/

Lepik, T. (2025, September 30). Näita, ära räägi [Show, don’t tell]. Blue Cat. https://bluecat.ee/en-blog-post.php?slug=naita-ara-raagi

Ling, A. (2016, December 12). Studio Ghibli's Hayao Miyazaki calls AI generated animation 'an insult to life itself'. We Are Resonate. https://www.weareresonate.com/2016/12/studio-ghibli-hayao-miyazaki-calls-ai-generated-animation-insult-life

Manovich, L. (2001). The language of new media. MIT Press.

Manovich, L. (2017). Automating aesthetics: Artificial intelligence and image culture. Flash Art International, 316. https://manovich.net/index.php/projects/automating-aesthetics-artificial-intelligence-and-image-culture

Manovich, L. (2018). AI aesthetics. Strelka Press. https://manovich.net/index.php/projects/ai-aesthetics

Manovich, L., & Arielli, E. (2024). Artificial aesthetics: Generative AI, art and visual media. https://manovich.net/index.php/projects/artificial-aesthetics

McLuhan, M. (1964). Understanding media: The extensions of man. McGraw-Hill.

Mori, M. (2012). The uncanny valley (K. F. MacDorman & N. Kageki, Trans.). IEEE Robotics & Automation Magazine, 19(2), 98–100. https://doi.org/10.1109/MRA.2012.2192811 (Original: Bukimi no tani. Energy, 7(4), 33–35, 1970.)

NielsenIQ. (2024, December 12). NIQ research uncovers hidden consumer attitudes toward AI-generated ads [Press release]. https://nielseniq.com/global/en/news-center/2024/niq-research-uncovers-hidden-consumer-attitudes-toward-ai-generated-ads/

Saar, E. (2015). Hirm tuttavliku ees: rännak õõvaorgu [Fear of the familiar: A journey into the uncanny valley]. Acta Semiotica Estica, 12, 68–84. https://ceeol.com/search/article-detail?id=535266

U.S. Copyright Office. (2025). Copyright and artificial intelligence, part 2: Copyrightability. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf