The Beatles are back, with a little help from their AI friends

On 2 November 2023, the world was treated to a special surprise: the release of a new Beatles song, Now and Then, made possible thanks to artificial intelligence (AI).

AI was used to isolate and restore John Lennon’s vocals from a previously unreleased demo recording, recorded by Lennon in 1979. It had been deemed unusable due to poor sound quality, but AI was able to clean up the recording and extract Lennon’s vocals, which were then combined with new instrumental parts recorded by Paul McCartney and Ringo Starr.

The official web site of The Beatles declared it to be the last Beatles song: “Together and apart, The Beatles have always had a talent for the unexpected. And now, 2023 brings one of the most anticipated releases of their long and endlessly eventful history. Now And Then is the last Beatles song – written and sung by John Lennon, developed and worked on by Paul McCartney, George Harrison and Ringo Starr, and now finally finished by Paul and Ringo over four decades later.

Now And Then was released worldwide by Apple Corps, Capitol and UMe. The double A-side single pairs the last Beatles song with the first: the band’s 1962 debut UK single, Love Me Do, which is described as “a truly fitting full-circle counterpart to Now And Then”.

Both songs are mixed in stereo and Dolby Atmos, and the release features original cover art by renowned artist Ed Ruscha. The new music video for Now And Then debuts today, November 3.

The release of Now and Then is a remarkable achievement, and it demonstrates the potential of AI to revolutionise the music industry. By making it possible to restore and remaster old recordings, AI can help to preserve musical heritage and make it accessible to new generations of fans.

How AI Made Now and Then Possible

The AI technology used to create Now and Then was developed by Peter Jackson, the director of the award-winning Beatles documentary series Get Back. Jackson’s team used a combination of machine learning and deep learning techniques to isolate Lennon’s vocals from the demo recording.

The first step was to train the AI to identify Lennon’s voice. This was done by feeding it a large dataset of Lennon’s vocals from other Beatles recordings. Once the AI was able to reliably identify Lennon’s voice, it was used to separate his vocals from the other instruments and noise on the demo recording.

The next step was to restore Lennon’s vocals. The AI was used to remove any noise or distortion from the recording, and to improve the overall sound quality.

Finally, the AI was used to combine Lennon’s vocals with new instrumental parts recorded by McCartney and Starr. The AI was used to match the tempo and pitch of the new instrumental parts to Lennon’s vocals, and to create a seamless blend of the two.

The release of Now and Then is just the beginning. AI has the potential to revolutionise the music industry in many ways. For example, AI can be used to:

* Restore and remaster old recordings

* Create new music and remixes

* Personalise music recommendations for listeners

* Assist songwriters and producers in the creative process