“I’m constantly trying to convince myself to remain optimistic about the future”: Inside South Korean entertainment's increasing reliance on AI
A troubled South Korean entertainment industry looks to AI for survival. If AI is the future, however, the broader ramifications remain unclear.
Outright resistant to AI-generated entertainment in South Korea is diminishing. At least according to Cho Kyung-hun, CEO of animation company Studio Animal. In places, that’s “simply familiarity,” he says. “Part of it might be a sense of resignation as AI becomes unavoidable.”
“The emotional climate is mixed and often more cautious,” says Jung Han, CEO of content studio K-EnterTech Hub (Korea-entertainment technology Hub). “Visible enthusiasm does not mean simple or uncritical acceptance.”
The tension is felt across many Korean studios I spoke to at the end of 2025 and into early-2026. And yet, forms of generative AI are being adopted at nearly every stage of entertainment production. And it may be a preview of what’s being used in US productions — if not already.
—
South Korea’s exploration of generative AI has been going on for years, but has been driven into hyperspeed by the new government’s intent to become a world AI leader. It’s made a commitment of roughly $1 billion to grow Korea’s animation sector specifically — through “AI-powered spin-offs” to existing properties, AI training programs, and resources to develop AI-generated virtual humans.
AI is already essential to expanding engagement with Korean content overseas: in recommendation algorithms, for localization, and dubbing. Dubbing is also increasingly valuable for Korea’s robust content creation industry. According to Han, this is supplemented by a strong virtual-human ecosystem, in which AI-generated avatars summarise the news and a broader idol culture supplemented by AI-generated influencers and popstars. Broadcast networks are leveraging that acceptance to generate narration over social media series summaries, which are themselves put together using AI scene selection and editing.
In more traditional entertainment, Korea’s largest television provider KT has developed models based on its 13 million subscribers, through which it analyzes billions of points of viewing data to identify what dramas and plots are likely to resonate most with viewers. This data, Han says, is “increasingly central to green-lighting decisions.”
Then there’s experiments in fully AI-generated content. Korean entertainment giant CJ ENM, the media company behind global hits including Reply 1988 and the Oscar-winning Parasite, released Cat Biggie in 2024—30 AI-generated 2-minute episodes of children’s content, developed by a team of six. Last year, a team of four at CJ ENM produced M Hotel, a 6-minute AI-generated film about an unhoused man who finds a key to a magical hotel.
Whether these experiments represent a harbinger of what’s to come is uncertain. CJ ENM declined to comment for this article. According to one studio head who wished to remain anonymous, however, if these experiments are perceived a success it could signal that large-scale investors intend to divest from smaller studios, towards AI.
A precarious financial situation has allowed AI to become so prevalent in Korean entertainment. If you view South Korean media solely through booming hits like Squid Game, you might be surprised that “structurally, it still sits in a relatively weak position,” Han says. Streaming has made Korean media accessible, but it remains far from a default choice for global viewers and broadcasters and the industry is desperate for ways to change, as series orders fall.
“The platform revolution has seen a significant shift in viewing habits [in Korea],” says Jason Bechervaise, an assistant professor specialising in Korean media and cultural industry at Hanyang University in Seoul. The change is shrinking the Korean movie industry too, with ticket sales plummeting from 200 million-a-year in 2019 to 106 million in 2025; “a 53% drop,” he adds, while Korean films only accounted for 41% of releases.
Multiple studios agree that the industry already lacks a sustainable model and faces chronic capital shortages. Desperation for solutions makes AI’s perceived potential impossible to ignore.
“Korean production studios do not welcome AI technology,” says Cho. “Studios are focused on survival—integrating AI into existing workflows, restructuring their organizations, and reshaping their business models to stay viable in this new environment.”
—
“In South Korea’s animation industry, subcontracting accounts for the overwhelming majority of work,” says Park In-chan, head of Yeson Entertainment’s digital team, a studio which provides assistive and in-between animation, including to US series Bob’s Burgers and Family Guy. This has been built on decades of relying on subcontracted work helping animators to develop. Now, even that status quo is uncertain.
“In-between animation has long been threatened by automation,” says Kim Myun-po, CEO of MAA MOPICS, which has worked with multiple Japanese clients, including on The Boy and The Heron. He believes it will take some time for AI to reach the necessary quality to be completely integrated into current processes. Han offers, however, that “the economic forces that once favoured Korean service work also incentivise Western studios to experiment with AI instead of some external vendors.”
It’s one reason why Yeson feels compelled to explore how AI can support its workflow and compete in an unsustainable market. Park cites, specifically, the ability to generate multiple camera angles of a scene. “These results are not usable as-is,” he adds, “but from a time-saving perspective, their value is significant and cannot be ignored.”
For now, these international relationships may dictate how and where AI is used, Kim says, adding that so far clients aren’t asking for AI to be used in outsourced animation work. “If we receive such a request, we’ll try,” he continues, “but I think that will likely be the day we stop doing business with them. It’s like requesting a hamburger from an Indian curry restaurant.”
—
We may be seeing the first hints that generative AI deployed to US-based projects through studios’ outsourcing work. “Assuming subcontracted production,” Park says, “I believe generative AI is already being partially used in western projects.”
For now, that appears limited to the same AI applications explored by Korean studios. Ironically, the in-between animation that once sustained Korean animation is also something Korean studios are looking to automate themselves, alongside a slew of applications they insist are “assistive”: cleaning up storyboards, converting 2D models to 3D, turning text into script format, creating design sheets, and offering variations on artwork. In many instances, these are generating references. Some studios, however, are already using AI to generate concept art to direct productions’ art-styles and in narrative considerations.
“By feeding in context such as character traits, personalities, behavioural patterns, and the key events outlined in the treatment,” Cho says, “we can use AI to examine multiple possible ways a character might react or make decisions within a given situation.”
These experiments are key, for some, to granting Korean entertainment more autonomy within the global market—and from the uncertainty of Korea’s subcontracting ecosystem—by turning to original content (and perceived copyright ownership). They see adoption and expansion of AI as increasingly vital.
“Competing through low-cost labor or simple production services is no longer a viable survival strategy,” says Hwang Jun-ho, CEO of Studio Popcorn. “We see AI as a way out of that structure. Shifting from a labour-centred to IP-centered production will be crucial for Korean animation to stand independently.”
Studio Popcorn is following through, too, releasing its own short animation series—26 7-minute episodes — Mongle Mongle Bakery, in 2026. It employs a hybrid production system with AI, with a reported 80% of elements in the series produced with AI. “As video generation models become more stable and controllable,” Hwang adds, “we believe that within the next two-to-three years, feature-length animation largely supported by AI will be possible.”
For Mofac, a VFX service studio that has shifted to a “content company”—and which experienced recent success with King of Kings — the future is even closer. “We are aware,” says CEO Jang Seong-ho, “that commercially-oriented feature-length content created solely using AI is already in production and prepared to make an appearance.”
They already have. I’m Popo, a fully AI-generated film, opened in May in the wake of CJ ENM’s The House, a hybrid of real actors playing against AI-generated environments. CJ ENM calls it a “new production paradigm.” Critics, however, are unimpressed. Though, according to Jung Han, this ignores the point of AI experimentation, which extends beyond box office numbers.
“The performance of any given title, regardless of how it was produced, is now tied to maximising the return on invested capital,” he says, across a portfolio of theatrical and streaming releases, “which is the crucial point for the industry right now.“
Not every studio is optimistic about the results, however, often echoing critics’ issues with AI-generated media in South Korean entertainment. “Whenever the material created by generative AI appeared on screen, it felt awkward and unnatural,” says Geum Dong-ho, founder and CEO of Seoul-based animation studio, Goldframe. When Goldframe experimented with generating assets, most had to be replaced by manual work.
Hwang acknowledges these limitations, referring to AI as a “highly-advanced brush,” inextricable from human input. He insists that AI and the small-scale productions it can support are essential to the future of Korean entertainment, but, in a refrain common across studios to whom I spoke, human creativity remains primary.
“There is a Korean proverb,” he offers, “Hwa-ryong-jeom-jeong (畫龍點睛), meaning ‘to add the final touch that brings life to a painting.’ AI can speed up production and reduce costs, but giving true life to a story will always remain the role of humans.”
JC Park, concept artist at Krafton, cites the same proverb as he urges caution. Though he suggests AI has potential for inspiration and reference, “the creative soul of a project isn't just a final layer of polish; it is the intent that drives every step,” he says. “Automating the entire process and only adding a human touch at the end might produce a functional product, but it risks losing the higher concept that connects with an audience.”
—
Studios risk more than a possible disconnect with audiences. Though the Korean government is committed to AI, the Supreme Court and National Assembly are yet to legislate on its use and training as they pertain to copyright and labour protections. Meanwhile, Korean broadcasters and rights holders are already litigating to stop their data being scraped.
Jang acknowledges that concerns around what many agree is a significant investment bubble around AI are founded. “Expectations are running ahead of reality,” he says, “and investment in future potential is overheating relative to the technology’s actual value.”
Studios maintain, however, that the technology is so powerful, so useful, and already so prevalent that adopters are unlikely to feel any great impact should the bubble burst or precedents turn against AI.
“It’s difficult for creators or studios to focus too much on the negative implications, the pace of change simply leaves little room for it,” Cho adds. “By integrating AI into government services, manufacturing, finance, healthcare, and content creation workflows, we are already seeing significant gains in efficiency and productivity. Those benefits won’t disappear even if investment trends cool.” This, despite even some of AI’s biggest proponents agreeing that few would be immune in AI-laced economies like the US and, now, South Korea should AI investment collapse.
In pushing AI harder despite the risks, The Korean government, Jang suggests, isn’t setting an agenda but rather responding to the demands of the industry while watching foreign law closely. Continued hesitation is unproductive.
“Many creators hold negative views [of AI] perhaps out of habit or past convention, the broader trend can no longer be ignored,” Jang says. “When silent films gave way to talkies, those who refused to adapt were naturally left behind. What we are facing today is a fear of the future rooted in uncertainty, along with an exaggerated sense of crisis driven by over-interpretation.”
Every studio speaking with me welcomed the idea of more solid legislation, agreeing that not only is clarity preferred, but vital.
“We are extremely cautious about copyright and data ethics,” Hwang says. “We avoid using datasets with unclear origins and rely only on in-house data or legally-secured materials for training and testing. Reasonable regulation will ultimately stabilise the industry and build trust.”
Hwang cites two models Studio Popcorn currently employs under this baseline. The first, Morphic AI, with which it has a partnership and, Jimeng AI.
Morphic did not respond to requests for comment about how its model is trained. It is clear about its origins, however. It’s based on Wan2.2, a model “trained on a vast dataset comprising billions of images and videos”—which can only be secured through public data — and including Koala-36M, compiled by scraping YouTube. Similarly, Jimeng AI is developed by a subsidiary of TikTok and Douyin owner ByteDance. Analysts and leaks agree that data from both sites is informing its datasets, as well as piggybacking off OpenAI and Microsoft Azure. Other studios cite Nano Banana Pro as widely-used, the image and video element of Google Gemini. Despite claims of legally-secured materials and desiring greater regulation, copyright-unsafe generation is already at the heart of South Korea’s AI efforts.
—
Having been a hub for subcontracted production, the future of Korean animation—and more broadly entertainment—is now uncertain. In an industry in which Kim says “working conditions and compensation are poor, and there are few good role-models for successful animation, leaving it lacking in promising career prospects,” the legality of the AI being deployed as an alternative to reform — whether through desperation, ignorance, or indifference — is just as shaky.
Any sense that a cycle of renewal might be imminent only feels farther off in an ageing industry — one struggling to attract young, entry-level applicants, while exploring automating the roles that would normally allow them to cut their teeth.
All of which, Cho suggests, only adds to an increasing “social pressure” within the industry to adopt AI before it’s left behind. Cho adds that it is necessary to regard AI as a tool (something we use) — rather than a technology (something in which we participate) — if only to justify having to use it to survive.
“I sometimes wonder if longing for a world without AI is simply too nostalgic,” he laughs. “I wish we could go back to a time before these technologies existed. In a way, I’m constantly trying to convince myself to remain optimistic about the future, even if that optimism doesn’t come naturally.”