Open DefinitlyEvil opened 6 years ago
and about inputs to the network, I think firstly everything should be transposed to a standard tonality, like C major or A minor(parallel modes). Then we should manually analyse all the harmonisations in the music piece and consider it as another vector.
I am a music college student majoring music composition and I am fond of polyphonic music. And I do code in serveral languages. @llSourcell is it possible to use both GAN networks and LSTM layers(like LSTM layers in both encoder and decoder)? and maybe do a video using Keras(that's the simplest lib I can understand). thx a looooot.