Jukebox

This artificial intelligence, named Jukebox, enables generating music, including rudimentary singing, in raw audio form across a variety of genres and artist styles. By providing the genre, the artist, and the lyrics as input, Jukebox produces a new musical sample entirely created from scratch. This model employs an innovative approach using an autoencoder to compress raw audio into a lower-dimensional space, then generate audio in that compressed space before resampling it to its original form.Musicians and producers can use Jukebox to experiment and create unique works, benefiting from quick adaptation to different musical styles thanks to training on pre-trained models from a vast database of 1.2 million songs. Not only does this technology facilitate musical creation, but it also opens the way to exploring new artistic dimensions through the social networks and other broadcasting platforms. The main advantage of Jukebox lies in its ability to produce music that respects the local coherence of musical sequences, while supporting music creators in experimenting with varied styles and integrating synthetic vocals. This represents a significant advance for those seeking to push the boundaries of creativity without the constraints of traditional workflows.