The idea of neural music is to:
- Convert an audio file to a lossless spectrogram
- Train a neural network to indentify patterns in a set of images (third party software can be used at this stage for example the deepdream from google)
- Generate music from noise