SEOUL, July 19 (Reuters) – In a dimly lit recording studio in Seoul, producers on the Okay-pop music label that introduced the world hit boy group BTS are utilizing synthetic intelligence to meld a South Korean singer’s voice with these of native audio system in 5 different languages.
The expertise enabled HYBE (352820.KS), South Korea’s largest music label, to launch a monitor by singer MIDNATT in six languages – Korean, English, Spanish, Chinese language, Japanese and Vietnamese in Might.
Some Okay-pop singers have launched songs in English and Japanese along with their native Korean, however making use of the brand new expertise for a simultaneous six-language launch is a worldwide first, based on HYBE, and will pave the way in which for it for use by extra fashionable acts.
“We might first hearken to the response, the voice of the followers, then resolve what our subsequent steps ought to be,” stated Chung Wooyong, the top of HYBE’s interactive media arm in an interview on the firm’s studio.
Lee Hyun, 40, generally known as MIDNATT, who speaks solely restricted English and Chinese language along with Korean, recorded the tune “Masquerade” in every language.
Native audio system learn out the lyrics, and later the 2 had been seamlessly mixed with the assistance of HYBE’s in-house AI music expertise, Chung stated.
The tune is the most recent signal of the rising affect of AI within the music trade at a time when the Grammy Awards have launched new guidelines for the expertise’s use and AI-generated mash-ups of songs are flooding social media.
“We divided a chunk of sound into completely different elements – pronunciation, timbre, pitch and quantity,” Chung stated. “We checked out pronunciation which is related to tongue motion and used our creativeness to see what sort of consequence we might make utilizing our expertise.”
In a before-and-after comparability proven to Reuters, an elongated vowel sound was added to the phrase “twisted” within the English lyrics, for instance, to sound extra pure whereas no detectable change was made to the singer’s voice.
[1/3]HYBE IM President Chung Wooyong speaks throughout an interview with Reuters at an organization workplace in Seoul, South Korea, June 22, 2023. REUTERS/Kim Hong-Ji
Utilizing deep studying powered by the Neural Evaluation and Synthesis (NANSY) framework developed by Supertone makes the tune sound extra pure than utilizing non-AI software program, Supertone chief working officer Choi Hee-doo stated.
HYBE introduced the 45 billion gained ($36 million) acquisition of Supertone in January. HYBE stated it deliberate to make a few of the AI expertise utilized in MIDNATT’s tune accessible to creators and the general public, however didn’t specify if it might cost charges.
‘IMMERSIVE EXPERIENCE’
MIDNATT stated utilizing AI had allowed him a “wider spectrum of creative expressions.”
“I really feel that the language barrier has been lifted and it is a lot simpler for world followers to have an immersive expertise with my music,” he stated in a press release.
Whereas the expertise just isn’t new, it’s an modern manner to make use of AI in music, stated Valerio Velardo, director of The Sound of AI, a Spain-based consulting service for AI music and audio.
Not solely skilled musicians but in addition a wider inhabitants will profit from AI music expertise in the long run, Velardo stated.
“It is going to decrease the barrier of music creation. It is slightly bit like Instagram for footage however within the case of music.”
For now, HYBE’s pronunciation correction expertise takes “weeks or months” to do its job however when the method quickens, it might serve a wider vary of functions comparable to decoding in video conferences, stated Choi Jin-woo, the producer of MIDNATT’s “Masquerade” who goes by the title Hitchhiker.
Reporting by Hyunsu Yim; Further reporting by Daewoung Kim and Hyun Younger Yi; Enhancing by Josh Smith and Jamie Freed
Our Requirements: The Thomson Reuters Belief Rules.