LSTM-music-gen

Description

This project aims to leverage the vast amount of MIDI files available on the web to train a Long Short-Term Memory (LSTM) neural network and generate new, original compositions. The core components of the project are a web scraper built with Selenium and an LSTM model constructed using Python.

Project Breakdown

1. The web-scraper (LSTM-music-gen/midi_extact_bot/midi_scrapper.ipynb) downloads the MIDI files from a predefined website into a designated folder (LSTM-music-gen/data/)
2. After training the model generates the new composition in accordance with specified model parameters and inputs (LSTM-music-gen/model/lstm.ipynb). The final outputs are populated in the respective directory (LSTM-music-gen/generated_output/).

A more detailed overview of the project is included in this repository's PowerPoint file (LSTM-music-gen/LSTM EDM Generation.pptx). This presentation provides in-depth insights into each phase of the project and can serve as a helpful guide to navigate through the project's progression.

Tech Stack

• Python (music21, midi2audio, keras, numpy, pickle, os, glob)
• Selenium (Chrome WebDriver)