The Beatles' AI-restored final song scores two Grammy nominations

Nearly five decades after breaking up, The Beatles are officially back in the Grammy race.

The legendary band earned two nominations for their final track, "Now and Then." The song is up for Record of the Year alongside heavyweight pop royalty like Beyoncé, Taylor Swift, Billie Eilish, and Charli XCX. It also picked up a nod for Best Rock Performance against Green Day, Pearl Jam, and The Black Keys.

Not bad for a track rescued from an old cassette tape using machine learning.

Unlocking a cassette tape from the 1970s

"Now and Then" originated as a home demo recorded by John Lennon in the late 1970s.

In the 1990s, Lennon's widow gave the tape to the three surviving Beatles as part of the Anthology project. They managed to finish other songs, but this one proved impossible. The analog technology of the era simply could not untangle Lennon's vocal from his background piano without degrading the sound into a muddy mess.

The breakthrough finally arrived in 2021. Director Peter Jackson and his sound production team developed machine learning tools capable of isolating vocals from instrumental backing tracks cleanly.

With that clean vocal stem finally in hand, Paul McCartney and Ringo Starr hit the studio to record fresh bass, drums, and backing vocals, officially putting the song together before releasing it in November 2023.

Staying on the right side of the rules

Despite the algorithmic assist, the track easily cleared the Recording Academy's rulebook.

Under current Grammy guidelines, only human creators are eligible to be nominated for or win an award. However, recordings that incorporate elements of AI material remain fully eligible across relevant categories.

The algorithm didn't generate any new notes or synthesize Lennon's voice. It simply functioned as a digital de-mixer, scrubbing away tape hiss and isolating a human performance that was recorded decades ago.

Now, the track will go head-to-head with modern anthems like Beyoncé's "Texas Hold 'Em" when the Grammy Awards air on Sunday, February 2nd, 2025.

Why it matters

This nomination sets a massive precedent for legacy artists, estates, and labels holding vaults of degraded analog recordings.

Up until recently, damaged demos and unmixed live tapes were essentially unusable commercially. By proving that source-separation machine learning can rescue archival audio up to Grammy-caliber fidelity, the industry just found a clear blueprint for catalog monetization without resorting to controversial generative mimicry.

We think the Recording Academy handled this smartly. While the music business spends enormous energy suing companies that train on copyrighted songs, recognizing "Now and Then" shows a clear distinction between generative replacement and technical restoration.

The catch, of course, is narrative weight. The Beatles pulled this off because hearing John Lennon sing across the decades carries unmatched emotional power. For lesser-known artists, cleaning up a dusty tape might yield a clearer file, but it won't automatically turn a forgotten outtake into a global phenomenon.