Developed by San Francisco-based research lab, OpenAI, the software is provided with genre, artist, and lyrics, and outputs a new music sample produced from scratch. Jukebox offers a selection of music styles, all in the vein of your favourite artists. For your listening pleasure, there’s rock in the style of Elvis, pop in the style of Céline Dion, and hip hop in the style of Kanye West, Tupac, and Nas. There’s also country, jazz, alternative metal, and even children’s music.
The tracks are split into four categories. In the first, the lyrics in each song are “co-written by a language model and OpenAI researchers”. Lyrics in a classic pop tune in the style of Frank Sinatra read: “It’s Christmas time, and you know what that means / Ohh, it’s hot tub time!”
Introducing Jukebox, a neural net that generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles. We're releasing a tool for everyone to explore the generated samples, as well as the model and code: https://t.co/EUq7hNZv62pic.twitter.com/sh5yHz7qrc
The second category is called ‘re-renditions’, in which Jukebox produces songs “very different from the original songs it was trained on”. The next is ‘completions’, in which researchers provide 12 seconds of audio, and Jukebox does the rest. One example in this category starts off as Mark Ronson and Bruno Mars’ “Uptown Funk” before the AI takes the track in a different direction, but still using the original lyrics.
In the final category, called ‘fun songs’, artificial intelligence takes one song and recreates it in the style of a different artist. For example, Lil Nas X’s “Old Town Road” in the style of country singer Alan Jackson, or “Lose Yourself” by Eminem in the style of Kanye – the mash-up you never knew you needed?