Gamers can chat in character voices with new Supertone AI

AI audio startup Supertone showcases two new voice generating technologies at the G-Star 2023 game festival, which runs at in Busan's Bexco convention center through Sunday. [HYBE]

 
BUSAN — Gamers, by as soon as next year, will be able to voice chat each other in the voices of their favorite characters in real time, using a new technology developed by AI audio startup Supertone.
 
Supertone, a subsidiary of the K-pop powerhouse HYBE that houses BTS, showcased two of its latest voice-generating AI technologies at G-Star 2023, which runs through Sunday.
 
Supertone, founded in 2020, has powered various audio stunts across K-pop and K-content and is now expanding to the game industry. Some of its gimmicks include manipulating the voice of Elvis Presley to sing a Korean singer’s ballad song and creating a brand new voice by blending those of actors Lee Han-byeol and Nana in the Netflix series “Mask Girl.”
 
The Korea JoongAng Daily tested Supertone's new technologies, dubbed Project Shift and Project Screenplay, at Busan's Bexco convention center on Friday. Both of them will be available for service next year. 
 
Project Shift is an audio service that analyzes a user's voice for 10 seconds as they read a provided text out loud. Based on the voice analysis, the tool can then convert the user's voice to those of 12 different proprietary game characters. 
 
When the journalist spoke into Supertone's microphone, a voice of a character named Arin from the upcoming role-playing game (RPG) ASTRA: Knights of Veda came out of the computer's speakers. The RPG will be published through HYBE IM, HYBE’s gaming subsidiary.
 
The voices can be adjusted to appropriately balance the user's voice with the character's voice, modify voice pitch and even control whisper levels, a function added to give more dimension to the vocals.
 

A visitor tries out Project Shift, a technology which can convert his voice to a game character's voice in real time at Bexco, Busan, on Friday. [HYBE]

 
“Right now, there are only 12 characters, but more will be added in the future,” said Yoo Han-kyul, director of Supertone’s business development unit. “This technology could be utilized across business-to-business [B2B] and business-to-consumer [B2C] sectors, such as when virtual K-pop groups communicate with their fans live, or when internet live streamers use it to immerse their subscribers to the games they play. It can also be used casually for entertainment, when gamers can voice chat in their characters’ voices while playing the game.”
 
Project Screenplay is a text-to-speech audio generating software for game developers. When a user inputs a dialogue into the software, a character’s voice can immediately read the text out loud. A voice’s pitch, enunciation speed, age and gender can all be modified accordingly.
 
“The generated voice can be freely distributed across the timeline and provide an all-in-one system for game developers from scriptwriting to dub the created voice to the game content,” Yoo said. “Project Screenplay can enhance the immersive experience within, as the flat, monotone voices of non-player characters or other characters can be replaced with voices generated by this service.”
 
Both technologies are currently available in Korean, English and Japanese.
 
If the two tools are released next year, the concept of voice licensing may also be put on the table for discussion, as iconic voices from human actors can also be deployed on for use their platforms.
 
“Right now, there is no legal act which incorporates voice licensing, as voice actors are paid to do their jobs on the spot,” a HYBE spokesperson said. “These technologies can be utilized to license the voice of an actor who depicted an iconic character in game or animation series, so symbolic that fans cannot imagine any other voice portraying that character. With the two technologies, that voice can be preserved forever to entertain the fans.”
 
Their path to profitability has not yet been determined, according to the HYBE spokesperson. 
 

BY LEE JAE-LIM [lee.jaelim@joongang.co.kr]