Learning Jazz Pianist Style with Cross-Attention Conditioning

A researcher has proposed a novel method for generating jazz piano music using a neural network architecture that incorporates cross-attention conditioning. The model is designed to learn the style of a specific jazz pianist by conditioning on their musical features. The architecture includes an encoder, a decoder, and a cross-attention mechanism. The encoder takes in a musical sequence, the decoder generates a new musical sequence based on the encoded features, and the cross-attention mechanism allows the decoder to attend to specific parts of the input sequence while generating the output. The model is able to generate coherent and style-consistent jazz piano music, and the author has released a demo implementation on GitHub.