Here you can find implemented RNN that composes music.
For training this RNN I used samples in boogie-woogie style. You can find it in /Samples directory.
You can check generated samples here: https://drive.google.com/open?id=1Ny8rjikoj6Xp7_9eKGzicqCz4fMV-9Bq
Or try to create music by yourself!
For using it the following libraries need to be installed:
numpy
, midi2audio
, mido
, torch
, os
(You can download them using pip3 install name_of_library
).
After all libraries installed:
-
clone repository
-
run
python3 run.py
from Sonia directory -
enter any integer number
-
now you can find 2 files in .midi and .wav format at current directory
This project was implemented basing on "This Time with Feeling: Learning Expressive Musical Performance" article (https://arxiv.org/abs/1808.03715).