A neural net that generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles.

JukeBox is a Neural Network that generates music, a project, realized by the OpenAI team. They developed the neural framework and trained it on 1.2 million songs and music pieces by various musicians, composers, and bands.

JukeBox uses a transformer but has another approach: sampling — and upsampling — of audio data. It does mean, this music model “knows” how various styles and voices sound — and writes its interpretations. As audio files.

Source: https://towardsdatascience.com/jukebox-by-openai-2f73638b3b73

Disclaimer: name and logo are the property of OpenAI

Icon by Ben Barry