Musenet

MuseNet is a powerful tool for creating musical compositions with a wide range of instruments and styles. It utilizes the same technology as GPT-2, which is a transformer model trained to predict the next token in a sequence. The model is trained on sequential data, and uses chordwise encoding to assign a token to each combination of notes sounding at one time. Additionally, composer and instrumentation tokens are used to give more control over the samples it generates. With this versatile system, MuseNet can synthesize music blending different genres while also remembering long-term structure within pieces.

The training dataset used by MuseNet is collected from various sources such as Classical Archives and BitMidi, as well as the MAESTRO dataset. The model has been tested on 4-minute long compositions with up to 10 instruments playing simultaneously, showcasing how far deep learning has come in producing human-like music in an automated way. This opens up interesting possibilities for artists and composers looking to explore new sounds or quickly generate unique pieces of music that blend multiple genres together.