Jukebox

Jukebox Overview

Jukebox is an innovative tool developed by OpenAI that uses genre, artist, and lyrics as input to generate new music samples from scratch. It utilizes an autoencoder to compress raw audio into a lower-dimensional space, discarding some of the perceptually irrelevant bits of information. The model then generates audio in this compressed space and upsamples back to the raw audio space. Jukebox is trained on a dataset of 1.2 million songs, paired with corresponding lyrics and metadata. It also allows for conditioning on artist, genre, and lyrics to generate music in a specific style.

Jukebox Highlights

• Jukebox uses an innovative approach to generate music directly from raw audio, overcoming the limitations of symbolic generators.
• The tool is trained on a diverse dataset of 1.2 million songs, paired with corresponding lyrics and metadata, allowing it to generate a wide range of music styles.
• Jukebox allows for conditioning on artist, genre, and lyrics, providing the ability to generate music in a specific style.