Jukebox

Jukebox is a neural network tool that is open-source and can generate music and basic singing in various genres and artist styles. It provides the code and model weights along with an exploration tool for the generated samples. Users can input their preferences for genre, artist, and lyrics, and Jukebox will produce new music samples accordingly. It is capable of producing a wide range of music styles and can even generate music that is completely different from the training songs when given lyrics that were part of the training data. With Jukebox, users can input 12 seconds of audio and the tool will complete the rest in a specified style. Jukebox tackles the challenge of working with long raw audio sequences by using an autoencoder to compress the audio to a lower-dimensional space. This allows the tool to generate audio in that compressed space and then up-sample it back to the raw audio space. Jukebox is an example of pushing the boundaries of generative models and is more expressive than tools that generate music symbolically. It is particularly suitable for users who want to experiment with AI-generated music.