AI Case Study

Sony CSL leads music collaborative Skygge to compose a music album with artificial intelligence

Researcher Francois Pachet at Sony Computer Science Laboratories (Sony CSL Paris) and Marie Curie University (UPMC) has created a music album composed by artificial intelligence. The album, called 'Hello World', was developed by artist or music collaborative, Skygge, as part of a project that explores AI in pop music production. The technology behind it is Sony's Flow Machines that runs on Markov chains. The system is trained on previously recorded music and analyses patterns to generate new melodies, accompaniments and instrumentation.

Industry

Consumer Goods And Services

Entertainment And Sports

Project Overview

"Hello World is the first album composed by an artist – SKYGGE – with artificial intelligence. Its goal is to show that AI can be used to create new, compelling music, generating fresh musical material of all sorts: melodies, harmonies, timbre, rhythms and the like.

Hello World started as a research project (the Flow Machines project) in which scientists were looking for algorithms to capture and reproduce the concept of musical “style”. Many scientific and technical results were obtained. Some prototypes were built with rudimentary interfaces (and a lot of bugs). The novelty and huge potential of the approach triggered the attention of a few talented musicians who joined the team. In 2016, we launched a song in the style of The Beatles, Daddy’s Car, as well as a less conventional title (Mr. Shadow) that would become the theme of this album. Then more musicians came. At some point, under the artistic direction of SKYGGE, the artists took control, and the scientific project became a music project.

This album is the result of that story. 15 songs were created by artists using Flow Machines: composers, singers, musicians, producers, and sound engineers, in many musical genres (pop, electronic, ambient, and jazz). With this diversity of skills, we had a single objective: use these new technologies to create novel, interesting music, yet music that would please our ears. Most importantly, music that the artists wanted to make! Melodic twists, harmonic surprises, and timbral juxtapositions, with a strong sense of direction and the uncompromising goal of making really good music, music that can touch and challenge the fan base of the artists.

The research leading to these results has received funding from the European Research Council under the European Union’s Seventy Framework Programme (EP7/2007-2013 Grant Agreement no.291156) and has been conducted by Francois Pachet at Sony Computer Science Laboratories(Sony CSL Paris) and Pierre and Marie Curie University (UPMC).
Flow Machines was developed by Vincent Degroote, Marco Marchini, Daniel Martin, Timotée Neullas, Alexandre Papadopoulos, Mathieu Ramona, Pierre Roy and Jason Sakellariou under the direction of François Pachet." (helloworldalbum.net)

"Flow Machines uses a probability equation, known as Markov chains to create catchy tracks.

Based on the information imputed and based on previously recorded music, Flow-Machines suggests melodies, accompaniments and instrumentation. Producers can accept, reject and alter these suggestions to create their AI-human collaboration." (phys.org)

Reported Results

A music album with 15 albums was produced.

Technology

"Flow Machines uses a probability equation, known as Markov chains to create catchy tracks"

Function

R And D

Product Development

Background

"Musical eras are often defined by their dominant modes of production—analog, electronic, digital—each bringing about new styles and ways of listening." (phys.org)

"In 1958, AI was used to compose Bach-like chorales (the Illiac suite). Since, huge progress has been made in AI technologies, including in machine-learning, combinatorial optimization, statistical inference, and related areas." (helloworldalbum.net)

Benefits

Data

Recorded music