Now and Then: Here’s how machine learning helped create The Beatles’ final song

Beatlemania 2.0

Now and Then: Here’s how machine learning helped create The Beatles’ final song

Today is a rather special day, as it marks the release of the first new song of The Beatles, which also happens to be their final song, since 1995. The song, Now and Then, has naturally drawn the attention of rock connoisseurs the world over.

After all, The Beatles happen to be one of the most influential rock bands of all time. This time, their song has caught the attention of techies around the world. Why, you ask? This is because Artificial Intelligence (AI) and Machine Learning (ML) happen to be perhaps, the only reasons why this release was possible in the first place. Intrigued? Read on.

How “Now and Then” was created

Paul McCartney and Ringo Starr, now the only two surviving members of the four-member band, had first gotten back together, along with George Harrison, after The Beatles split up in 1974. Two unreleased songs that Lennon had recorded in his voice – Free as a Bird and Real Love were completed by layering instruments on top.

The third song – Now and Then – could not be completed. This is because Lennon’s vocals on it, were often faded at certain points, suppressed in others, due to the piano.

ALSO READ: From Vinyl to Virtual: How gadgets have mirrored and shaped the way we listen to music

However, while working on the documentary Get Back, with Peter Jackson in 2022, they discovered and developed a machine learning technique to isolate all the different instruments and voices in a song, each into a separate track.

The model could then be trained on multiple tracks and learn the waveforms of each individual sound in them. Following this, the model was able to isolate each of these individual sounds and extract them, even when multiple different tracks are merged into as single one.

This technology, however, isn’t entirely unheard of. It is commonly used for phone calls now as well, using models trained on human voices. What it does is essentially suppresses any voices that aren’t the speakers’, such as traffic on the road or speakers playing music loudly on the street.

While one may question the song’s authenticity, it is important to understand that it isn’t AI-generated. AI over here, was simply used to isolate Lennon’s voice such that it could be heard clearly, while original music, with Starr on the drums, was added to it.

Now and Then is now available to stream on YouTube, Spotify, Prime Music, as well as Apple Music.

Unleash your inner geek with Croma Unboxed

Subscribe now to stay ahead with the latest articles and updates

You are almost there

Enter your details to subscribe

0

Disclaimer: This post as well as the layout and design on this website are protected under Indian intellectual property laws, including the Copyright Act, 1957 and the Trade Marks Act, 1999 and is the property of Infiniti Retail Limited (Croma). Using, copying (in full or in part), adapting or altering this post or any other material from Croma’s website is expressly prohibited without prior written permission from Croma. For permission to use the content on the Croma’s website, please connect on contactunboxed@croma.com

Comments

Leave a Reply
  • Related articles
  • Popular articles
  • Audio

    Deepfake audio is not so easy to detect

    Atreya Raghavan

  • Laptops

    ChatGPT gets support for voice and image prompts

    Atreya Raghavan

  • Smartphones

    Apple may be foraying into AI

    Atreya Raghavan