He may have died 25 years ago, but one of South Korea's biggest rock legends, Kim Kwang-seok, is performing again…
with the help of artificial intelligence.
AI company Supertone has resurrected Kim’s voice for a new South Korean television show airing on SBS.
Successfully getting the AI voice to express human emotion in its songs was an important goal, according to one of the show’s producers Kim Min-Ji.
"We were very surprised to hear that the AI could sing better after doing a vocal warm up, like 'A-E-I-O-U'. It is even more interesting that it was a big surprise to the developers as well. It's same with a human being who can sing better when his or her throat is warmed up."
The series showcases what’s called ‘’Singing Voice Synthesis’’ technology, which essentially mimics the range of human vocal cords when provided with an original to work from – in this case, Kim’s voice.
To start with, the program learned 20 of the late singer’s songs.
Technicians then trained the AI system to sing over 700 Korean songs with the same voice.
While some fans cherished the opportunity to hear the singer’s voice once again, the producers admitted that there were some initial ethical concerns surrounding the technology - not to mention the potential copyright issues.
"What we were most scared and worried about was asking Kim's family for their permission. Kim Kwang-seok has a huge fan base, so we were also very worried about possible backlash. But when we played the AI audio to his family, they were very happy with it. They were wary at first, but after listening to the outcome, they said it was like Kim Kwang-seok came back alive."