LONDON (AP) – Decades after the band broke up, artificial intelligence was used to extract John Lennon’s voice from an old demo to create ‘the last Beatles record,’ Paul McCartney said Tuesday. announced to
McCartney, 80, told the BBC that the technology was used to separate the Beatles’ voices from background sounds during the production of Peter Jackson’s 2021 documentary series The Beatles: Get Back. said. The “new” song is expected to be released later this year.
Jackson “was able to pull John’s voice out of a little ropey cassette and piano,” McCartney told BBC radio. “He can use AI to separate them out and tell the machine, ‘This is the voice, this is the guitar, get rid of the guitar.'”
“So when it came down to making what would be the Beatles’ last record, it was a demo that John had and we worked on it,” he added. “We were able to extract John’s voice purely through this AI, so we could mix the record just like you do. It gives us some leeway.” .”
McCartney did not name the song on the demo, but the BBC and others said it was likely John Lennon’s unfinished 1978 love song “Now and then.” The BBC reports that the demo was on a cassette labeled ‘For Paul’ that McCartney received from John Lennon’s widow, Yoko Ono.
McCartney called AI technology “in some ways scary, but exciting,” adding, “We’ll have to see where it goes.”
The same technology allowed McCartney to virtually “duet” “I’ve Got a Feeling” with the 1980-murdered John Lennon at Glastonbury Festival last year..
Holly Herndon, a multidisciplinary artist with a PhD in composition from Stanford University, used early AI machine technology on her final album Proto in 2019, Holly+, an online protocol that allows members of the public to upload tracks to be reinterpreted and performed in a deepfake version of her voice. She theorizes that the Beatles recordings were likely made using a process called “source separation”.
“With machine learning, source isolation has become much easier, allowing us to extract and isolate audio from recordings and add new instruments,” she explains.
it’s different from deepfake vocals. “Deepfakes are brand new vocal lines created from machine learning models trained on old vocal lines,” she said. “It doesn’t seem to happen in this example, but the analysis of old material has allowed us to create endless new media. It’s a similar process in spirit to this song.”
McCartney plans to open an exhibition at London’s National Portrait Gallery later this month, showcasing never-before-seen photographs taken during the early years of Beatlemania, the time the Beatles gained worldwide fame.
Entitled Eye of the Storm, the exhibition features more than 250 photographs taken with McCartney’s camera from 1963 to 1964, including Ringo Starr, George Harrison, Also included are portraits of Lennon and Beatles manager Brian Epstein.
___
This article has been corrected to indicate that McCartney’s photo exhibition is titled “Eyes of the Storm” instead of “Eye of the Storm”.
___
Sherman reported from Los Angeles.
