July 2, 2025
SEOUL – “Create a video of a young Korean girl playing at a beach.”
From the way the girl cautiously steps into the water to the sunlight glimmering on the ocean surface, it was remarkable how much detail the video managed to capture from such a simple prompt.
This prompt-based AI video also features realistic sound effects, such as crashing waves and the gentle splash of water with each step she takes.
Amid rapid advances in artificial intelligence, including the May 20 release of Google’s Veo 3, content created using AI tools has taken over social media, ranging from “mukbang” eating shows to autonomous sensory meridian response, or ASMR, videos.
On TikTok, for example, AI-generated videos of glass fruits and vegetables being sliced or eaten have gone viral at a rapid pace, inspiring a wave of similar content from other creators.
AI-generated videos on TikTok’s WildWhispersASMR channel featuring glass-like fruits and vegetables being sliced or eaten have gone viral, garnering millions of views. PHOTO: WILD WHISPERS ASMR/THE KOREA HERALD
A video uploaded by the channel WildWhispersASMR on June 13, showing a glass kiwi being sliced, received 68 million views in just eight days. The crisp, satisfying sound of a knife or teeth hitting the glass surface delivers a unique sensory satisfaction that keeps viewers hooked.
AI video generation tools like Veo 3 and OpenAI’s Sora, which was released late last year, are now widely used not only by video professionals but also by ordinary users, with more people branding themselves as AI video creators.
Son Min-yeong used to be an ordinary office worker with no experience in video editing. but thanks to AI tools, she has recently become a digital creator with over 20,000 followers on Instagram.
“Because of health issues, I started looking for ways to earn money online. But I wasn’t confident about showing my face or producing high-quality videos. It was AI video tools that helped me to enter the digital content market,” she told The Korea Herald.
“Generating high-quality videos with AI requires detailed prompts that include specific examples — it’s like having a deep conversation with the AI.”
What draws people to AI-created videos is the challenge of determining whether they’re real or not, fueling intrigue and fascination among viewers.
Yu Ha-seong, a 21-year-old university student in Daegu who follows creators of virtual mukbang content on Instagram, said, “I was shocked by how realistic it looked. One video even showed oil dripping from a chicken drumstick as a female creator took a bite.”
“When she was shown in close-up, even her skin, double eyelids, and stray baby hairs looked completely real,” he added.
A short AI-generated clip posted on Instagram features a virtual mukbang influencer eating jelly. PHOTO: INSTAGRAM/THE KOREA HERALD
The realistic reproduction of human speech makes the experience even more immersive.
AI video generation tools have long struggled to synchronize lip movements with audio, but the recently launched Veo 3 has made notable improvements in this area, producing visuals so convincing they can easily fool viewers.
In one video, a YouTuber appears to be interviewing a passerby on the street. From the YouTuber’s subtle facial expressions to the interviewee’s natural pauses before responding, every gesture closely mimics real human behavior. Even the outfits and the urban backdrop resemble a typical day in Seoul. Yet none of it is real. It’s all generated by AI.
Beyond replicating real-life scenarios like mukbang shows or street interviews, AI-generated clips portraying what would be impossible situations in reality have also captured viewers with moments of intense visual pleasure.
A recent viral video on Instagram features a middle-aged man from the Joseon era eating a bowl of freshly cooked rice topped with kimchi, mimicking the style of a modern mukbang YouTuber. The creative blend of historical settings and contemporary internet culture quickly drew widespread attention.
In an AI-generated clip on Instagram, a woman goes down a slide amid a slick of tteokbokki. PHOTO: INSTAGRAM/THE KOREA HERALD
Among other quirky, surreal examples are slides flowing with tteokbokki (spicy rice cakes), reporters interviewing dogs that speak Korean, and swimmers gliding through lava.
Creators often exchange ideas and tips for writing prompts via comment sections or private messages, helping one another produce more lifelike videos using AI video generation tools.
As AI video creation emerges as a new form of online play culture, concerns are also growing over its unintended side effects.
Some criticize digital creators for generating AI videos that feature women in bikinis or lingerie presenting news or mukbang content.
“With deepfakes already being such a big problem, I’m worried that these AI video tools will just make it worse,” an X user wrote.
Hwang Ji-yeon, 26, a freelance illustrator based in Seoul, said, “I once watched an AI-generated clip that presented false information about major political issues in the form of interviews, seemingly intended to promote the creator’s personal views. What scared me was how real the people looked. I felt that older people might easily be fooled.”
Lee Se-young, a professor of media communications at Sungkyunkwan University, echoed Hwang’s concerns, warning that the lower barriers to creating AI-generated videos could further entrench the “post-truth society,” where facts become less influential than emotions.
“The growing accessibility of AI tools to the general public brings about fresh creative possibilities, but also raises serious concerns over the spread of misinformation, such as fabricated interviews, altered quotes and reputational harm to public figures,” Lee told The Korea Herald.
“These developments not only mislead audiences but also threaten the credibility of the media itself, ultimately deepening the ‘post-truth’ society, where facts matter less than emotions.”
The professor stressed that clearly labeling AI content is essential for ensuring users can make informed choices.
“It reminds content creators of their ethical obligations and pushes platforms to be more cautious in handling AI-based material,” he explained.
Under Korea’s Basic Act on Artificial Intelligence, set to take effect in January, all content created using generative AI, including films and TV dramas, must carry a clear label indicating its AI origin.
Media users, the professor added, now require what he called a new form of media literacy: AI discernment.
“With technology advancing rapidly, viewers can no longer rely on visuals alone to assess videos’ authenticity. That’s why they need sharper analytical skills, from verifying the sources of videos and understanding how they were created using AI tools, to evaluating the cultural and social context behind the content.”
