The Beatles release final song with help from AI audio tech

The final Beatles song is officially here, and it took a mix of half-century-old cassette tape, 1995 archive sessions, and Peter Jackson’s machine learning tools to make it happen.

Running four minutes and eight seconds, “Now and Then” represents the first and only original Beatles recording of the 21st century. Released alongside the band's 1962 debut single “Love Me Do” as a double A-side single, the track features contributions from all four original members: John Lennon, Paul McCartney, George Harrison, and Ringo Starr.

Rescuing Lennon’s voice from a dusty demo

The origins trace back to the late 1970s, when Lennon recorded a home demo on a cassette tape at his Dakota apartment. In the mid-1990s, Yoko Ono gave the tape to McCartney, Harrison, and Starr. While they successfully polished off “Free As a Bird” and “Real Love,” “Now and Then” proved unworkable. Lennon's vocals were badly buried under his own piano playing, and the audio separation tech of that era just wasn't up to the task. After running out of steam, the tape went into a cupboard for nearly a quarter of a century.

Everything changed thanks to director Peter Jackson. While directing the 2021 Disney Plus documentary “The Beatles: Get Back,” his team developed machine learning software capable of isolating individual voices and instruments from murky vintage recordings. In 2022, McCartney and Starr realized they could run Lennon's original demo through that pipeline to extract his vocal track completely clean.

With Lennon’s voice isolated, McCartney recorded bass, Starr added drums, and the duo preserved guitar parts Harrison recorded during their 1995 sessions. McCartney even laid down a slide guitar solo in Harrison’s signature style. Giles Martin, son of legendary Beatles producer George Martin, crafted an ambitious string arrangement. To keep the project secret, McCartney told the session string players that they were recording for a solo project.

The Fab Four reunite on screen

Alongside the release, Apple Corps dropped a mini-documentary detailing the song’s creation, with Jackson directing an official music video set to debut Friday. Jackson’s visual treatment blends modern shots of McCartney and Starr with 14 hours of forgotten 1995 studio sessions, private home movies, and never-before-seen footage of the Beatles performing in leather suits provided by original drummer Pete Best.

Within 11 hours of hitting YouTube, “Now and Then” gathered over 3.4 million views and was named BBC Radio 1’s Hottest Record. Lennon’s son, Sean Ono Lennon, defended the process, noting his father loved experimenting with recording tech and would have embraced the project.

Why it matters

There is a massive distinction between generative AI creating artificial voice clones and machine learning acting as an audio scalpel. “Now and Then” is firmly the latter. McCartney and Starr did not synthesize Lennon's voice; they excavated his actual performance from tape hiss and a loud piano. For listeners, it delivers genuine historical closure rather than cheap synthetic nostalgia.

For the music industry, this release is a watershed moment for archival restoration. Jackson’s AI sound-separation tech proves that low-fidelity cassette demos, muddy bootlegs, and damaged analog tapes can be professionally remixed into modern Atmos-grade standards. We expect major labels to aggressively comb through vaults belonging to Nirvana, Jimi Hendrix, and David Bowie to restore lost recordings that were previously deemed unusable.

The only real risk here is boundary creep. The Beatles pulled this off with full blessing from Lennon and Harrison’s estates, genuine unreleased performances, and the surviving members at the mixing console. If lesser-scrupled labels adopt this technology without the same emotional reverence, we could quickly see an endless wave of half-baked posthumous releases flooding streaming platforms.