The Beatles release their “last” record. Artificial intelligence helped make this happen

Paul McCartney said on Tuesday that artificial intelligence was used to extract John Lennon’s voice from an old show to create “the last Beatles record,” decades after the band broke up.

McCartney, 80, told the BBC that the technique was used to separate the Beatles’ vocals from background sounds during the production of director Peter Jackson’s 2021 documentary series, “The Beatles: The Return”. He said the “new” single is set to be released later this year.

McCartney told BBC Radio that Jackson “was able to get John’s voice out of a little cassette tape and out of a piano.” “He can separate them with AI, he’ll say to the machine ‘This is a voice, this is a guitar, I lose the guitar.'”

He added, “So when we came in to do what would be the last Beatles record, it was a demo for John and we worked on it.” “We were able to take John’s voice and make it pure with this AI so we can mix the recording like you do. It kind of gives you some leeway.”

McCartney did not name the demo, but the BBC and others said it was likely an unfinished love song by Lennon on 1978’s “Now and Then”. The demo was included on a cassette tape titled “For Paul” that McCartney had received from Lennon’s widow, Yoko Ono, the BBC reported.

McCartney described the AI ​​technology as “kind of scary but exciting”, adding, “We’ll just have to see where that leads.”

The same technology enabled McCartney to “duet” virtually with Lennon, who was murdered in 1980, on last year’s “I Got a Feeling” at Glastonbury.

Holly Herndon, a multidisciplinary artist with a Ph.D. in installation from Stanford University, used emerging AI machine technology on her latest album, 2019’s “Proto,” and developed Holly +, an online protocol that allows audiences to upload audio clips to be reinterpreted and performed by a fake version of their own voice. I assumed the Beatles recording was created using a process called “source separation”.

“Separating the source is made much easier to deal with with machine learning. This allows you to extract audio from a recording, isolating it so that you can accompany it with new hardware,” she explains.

This is different from the deep voice. “Deepfake is a completely new voice font that was born from a machine learning model that was trained on old voice fonts,” she said. “While this does not appear to be happening in this example, it is now possible to produce infinite new media from the analysis of old material, a process similar, in spirit, to this song.”

McCartney is set to open an exhibition later this month at the National Portrait Gallery in London showcasing never-before-seen photographs he took during the Beatles’ early days at the onset of “Beatlemania,” when the band rose to international fame.

The exhibition, titled Eyes of the Storm, displays more than 250 photographs McCartney took with his camera between 1963 and 1964 – including those of Ringo Starr, George Harrison and Lennon, as well as of Beatles manager Brian Epstein.

___

This story has been corrected to clarify that the title of the McCartney portrait gallery is “Eyes of the Storm” and not “Eye of the Storm”.

___

Sherman reported from Los Angeles.