The scope of AI applications has been expanding rapidly in recent years and has begun to affect creative activities such as painting, literature, music, and dance. In the visual arts in particular, the generation of paintings using generative adversarial networks based on deep learning techniques has led to significant advances, and these generated works have even been sold at auctions for high prices. This white paper focuses on such technological developments in AI in the creativity domain, particularly in the area of music. Although relatively less talked about than image and text generation, this is an area of interest that is undergoing significant change thanks to recent developments in deep learning technology. In this white paper, we will discuss the use of AI in the music field including its technological trends, its business applications, and its impact on society.

Table of Contents1. Purpose of this Report
2. Types of AI Music Generation Systems
3. Application Areas and Business Potential
4. Social Implications and Challenges of Music Generation Technology
5. Development and Current Status of Music Generation Technology
– Music generation before deep learning
– AI (deep learning) development of music generation technology
– Future research issues
6. About the Author
7. Recommended Information Sources
8. References

Media, contents

In this section we present the multimedia contents of the white paper. Please download the entire white paper by using the button at the top.

Types of AI Music Generation Systems

Pop song accompaniment generated by PopMag

Songs generated by FIGARO

Frank Sinatra singing “Somewhere over the Rainbow”

Soundify finding the appropriate audio clip for various input videos

Application Areas and Business Potential

Daddy’s Car composed by Flow Machines

AI and Human Improvisation in the AI DJ Project

Perpetually Performing Death Metal by Databots

Development and Current Status of Music Generation Technology

Yanis Xenakis, “Metastasis.”

Piano music generated by Performance RNN

An example of a melody embedded in the latent space of MusicVAE

Interpolating a melody using MusicVAE

Ableton Live plug-in of GrooVAE

M4L.RhythmVAE – Rhythm generation device built for Max for Live/Ableton Live

Audio samples generated by MuseNet

An example of audio style transfer using MuseMorphose

An example of rain sounds generated by Word2Wave

Examples of video and sound matching using Soundify

NSynth Super Overview

Future research issues

Result of using DDSP to generate a cello sound based on a violin sound

Novel rhythms generated using CAN