marysalvana / sonia

Generating music using various approaches.

Sonia

Generation of music using autoencoders.

This repository contains all follow-up materials, such as .ipynb labs, Kaggle competitions submission notebooks, and other stuff that gives enough foundation for the project.

The project itself with everything required for its work lies in Project folder.

Run download_project.sh to download necessary files for project usage.
Run download_kaggle.sh to download datasets for Kaggle competitions.

---

Quick note on usage of the project:
Make sure that midi2audio and fluidsynth are installed (pip install midi2audio && sudo apt install fluidsynth)

1. python3 inference.py
2. Enter random seed, "density rate" (float number from 0 to 1, the lower - the denser, 0.25 is optimal for most cases), output directory and path to the pretrained model (mine are stored in States folder).
3. Generated music (.mid and .wav) would lie in the provided folder.