ledstarr / transcribed_music_data_generator

Synthetic Music Database Geneartor

Transcribed Music Data Generator

This Script is intended to help Musicologists to generate randomly generated MIDI and WAV Files.

First it will generate a MIDI File and then synthesizes a WAV File from it. Theses Files will be stored inside a CSV or a Train and Test CSV afterwards and stores the relative paths to the newly generated Files with some extra information.

This Data is intended to help in cases where transcribed Music-Data is scars and or the human bias is something you want to cancel out of the equation.

By running main.py the Script will ask you some questions and generate all Files afterwards. Feel free to use the batch_generate() Method from the DataBaseGenerator Class for creating the Data-Sets on your own.

Dependencies:

This script uses:

• music21 for creating the MIDI Files
• mido for reading the MIDI Files in a stream.
• pyo as sound engine for synthesizing the WAV Files
• pandas for creating the CSV's
• numpy for creating random numbers
• tqdm for displaying a progress bar

Since pyo only supports Python 3.7 it is currently not possible to use this script with higher Versions of Python.