OpenAI’s JukeBox is a Neural Network that creates music, from rudimentary singing to complete songs from various genres and artist styles. The project was developed by the OpenAI team who trained the neural framework on 1.2 million recordings of different musicians, composers, and bands. JukeBox uses a transformer approach but takes it one step further with audio data sampling and upsampling techniques in order to teach itself how each style and voice sounds like – then generating its own interpretation of these as audio files.