The music video for Eternity’s latest single “DTDTGMGN” showcases the band’s synchronised dance moves to a catchy bubblegum pop rhythm within a vibrant neon pink room adorned with school uniforms and celebrity posters. The camera’s dynamic movements and rapid scene transitions reflect the characteristic style of K-pop, a genre that has gained worldwide popularity in recent years.
The 11-member group Eternity was created by South Korean music management company, Pulse | Credit: Pulse9
As one member croons, “Hold me tight, lean on my shoulder,” fans of the group are unable to physically fulfill these desires. This is because the 11 members of Eternity are not real individuals but rather virtual entities.
Utilising AI technology, Eternity stands at the forefront of South Korean entertainment, blurring the line between reality and virtuality. Pulse9, the group’s creator and management company, firmly believes that computer-generated stars possess a distinct advantage over their real-life counterparts.
Similar to many popular K-pop acts, Eternity consists of a group leader, vocalists, rappers, and dancers. However, one member, Zae-in, stands out with her exceptional versatility. “Zae-in possesses a combination of talents that would challenge most human artists,” explained Park Ji-eun, CEO of the company. “She excels in singing, rapping, and even showcases skills as a fashion designer. Additionally, she is a naturally gifted actress.”
Utilising cutting-edge real-time face-swapping technology and an AI-generated voice, Zae-in is portrayed by a collective of 10 individuals, including actors, singers, and dancers, who contribute their talents to her digital persona as required. An example of this collaboration can be seen in a video shared on Zae-in’s Instagram account, where she effortlessly rides a longboard alongside Seoul’s Han River. In this particular video, the creators filmed the stunt with a stand-in whose face was subsequently replaced with Zae-in’s.
During a Zoom interview the virtual star expressed, “There are different versions of Zae-in, such as the pop music Zae-in, actress Zae-in, and reporter Zae-in.” It is worth noting that an actor portrayed Zae-in during the interview, although Pulse9 stated that their AI software can replicate Zae-in’s voice and generate speech independently. Zae-in elaborated further, stating, “The existence of various Zae-ins gives us a unique advantage, allowing us to accomplish what a single person cannot.”
She further commented, her eyes gleaming with a smile, “The only thing we are unable to do is sign autographs.”
Real-time face-swapping and early iterations of “deepfake” technology have been in existence for nearly a decade. During this time, virtual influencers such as Lil Miquela and South Korea’s own Rozy have amassed substantial online followings in the realms of fashion and music. However, recent advancements in AI-generated video and audio are simplifying the process for companies like Pulse9 to craft believable virtual characters. Moreover, progress in generative AI programs like ChatGPT is eliminating the requirement for individuals to “play” characters like Zae-in altogether.
The process of bringing Eternity to “life” commenced with the team at Pulse9 undertaking the task of designing the appearances of its members. Park’s team created 101 distinct digital faces, drawing inspiration from renowned Korean celebrities. These faces were subsequently categorised into four groups labeled as “sexy,” “cute,” “intelligent,” and “innocent.” Through an online poll, the selection was narrowed down to 11 faces.
Employing deepfake technology and motion capture techniques, Pulse9 proceeded to produce the band’s inaugural music video, which accompanied the ironically titled track “I’m Real,” in 2021. The video elicited mixed reactions from viewers. Some social media users felt that the band members’ facial expressions appeared unnatural and disconnected from their bodies, while others pointed out the disconcerting “uncanny valley” effect, which arises when human-like entities evoke feelings of unease.
However, five months later, when Pulse9 released Eternity’s subsequent music video, “No Filter,” followers observed that the group was nearly indistinguishable from real individuals. The most popular comment on YouTube reads, “I’m surprised by how realistic they sound, and the visuals are even more realistic than their debut song.”
The release of “DTDTGMGN” in October represented another significant leap in quality. Park stated, “Until last year, only the face was virtual, but today we can create entire full-body images using AI technology.” Additionally, Park mentioned that “No Filter” was composed using AI, although human intervention was involved in editing the song and writing the lyrics.
S M Entertainment, the powerhouse behind K-pop sensations like Super Junior, Girls’ Generation, and Exo, has also dabbled in the realm of AI. In 2020, the company introduced Aespa, a group consisting of four human artists and their virtual counterparts. Following the group’s debut, SM’s founder, Lee Soo-man, proclaimed that the future of entertainment lies in the fusion of “celebrity and robot.”
In a different vein, Hybe, the management company of BTS, utilizdd AI to refine the pronunciation and pitch of one of its singers before releasing a song in six languages: Korean, English, Spanish, Chinese, Japanese, and Vietnamese.
Earlier this year, Metaverse Entertainment, a collaboration between video game publisher Netmarble and Kakao Entertainment (the parent company of SM Entertainment), unveiled a virtual girl group called MAVE. Unlike Eternity, which relies on straightforward face-swapping techniques, MAVE’s four members were entirely generated from scratch using CGI.
Members of K-pop quartet MAVE can sing, dance and speak multiple languages – all with the help of Al | Credit: Metaverse Entertainment
AI technology was employed by Metaverse Entertainment to create facial features by merging the most desirable traits and hairstyles, resulting in the final characters. Motion capture technology was then utilised to record real-life K-pop performances, which were subsequently animated by a deep learning program for music videos. According to Kang Sung-ku, the chief technical officer of Metaverse Entertainment, the vocals of the group are a blend of human and AI-generated elements.
“Our objective is to develop fully AI-driven virtual humans that are genuinely believable,” Kang Sung-ku expressed in an interview with CNN.
MAVE’s debut single, “Pandora,” has gained an impressive 25 million views on YouTube since its January release.
Is this where K-pop is headed?
Virtual popstars might have an edge over human idols in terms of connecting with their fans. Advances in AI language technology could enable fans to communicate with their idols, fostering personal relationships. Kang, the creator of these AI idols, likened them to Siri, the virtual assistant on iPhones, noting their ability to remember previous conversations and engage based on that information.
Eternity’s Zae-in told CNN that virtual stars’ “unique advantage” is that they “can do what one person cannot.” | Credit: Pulse9
Metaverse Entertainment is already exploring this concept through apps that allow users to interact directly with MAVE. The band members currently speak multiple languages, including Korean, English, French, and Indonesian, reflecting their diverse fan bases and target markets. In theory, they could learn and communicate in any language.
Kang believes that having a personalised AI idol could provide emotional support and care that may not be possible with even close friends. This development could benefit both individuals and companies like Kang’s. The K-pop industry heavily relies on an arduous training system, where aspiring artists undergo years of expensive dance and music classes from a young age.
Agencies in the K-pop industry invest significant resources in training their artists, including years of expensive dance and music classes. For instance, SM Entertainment reportedly spent around $3 million to debut Girls Generation in 2007, covering costs such as lessons, studio space, styling, and even cosmetic procedures. These expenses are not necessary for virtual talents.
In addition, computer-generated pop stars have certain advantages over real-life idols. They don’t fall ill, age, or get involved in contract disputes or tabloid scandals. However, Kang acknowledges that major agencies are unlikely to replace their human bands any time soon. Some fans have reacted negatively to AI groups, fearing that real-life idols might lose their positions.
Kang explains that the criticism revolves around the idea of virtual idols replacing current ones. Nevertheless, there will always be fans who prefer human artists. Park, the creator of Eternity, shares a similar perspective and compares the emergence of AI in the industry to the invention of photography in the 19th century. Just as painters still exist today despite the existence of photography, BTS and other human idols have nothing to fear from virtual idols—they are simply new forms of content.
Subscribe to FIB’s Weekly Breaking News Report for your weekly dose of music, fashion and pop culture news!