Description

This repo helps prepare your data model set for generating a custom model for Audiocraft/Music Gen.

I recommended using these helpful scripts and then heading over to https://github.com/chavinlo/musicgen_trainer and using those scripts to build your model to import into AudioCraft/MusicGen.

Download Song Script

This script allows you to download a song from YouTube and convert it to .wav format.

Requirements

Install FFmpeg. The installation process varies depending on your operating system. You can find instructions for your specific OS on the official FFmpeg website.

Also run:

pip install -r requirements.txt

Usage

You can use the script from the command line by providing the YouTube URL and the output directory as arguments:

python download_song.py <youtube url> <output directory>

For example:

python download_song.py https://www.youtube.com/watch?v=k4jDpkAWrdA out

If you run the script without any arguments, it will prompt you to enter the YouTube URL and the output directory:

python download_song.py

You should now have a directory with .wav file/s

After downloading the .wav/s, you can now use split_song.py

python split_song.py <directory with .wav/s> <output directory>

For example:

python split_song.py out data_model

If you run the script without any arguments, it will prompt you to enter the directory with the .wavs and the output data_model directory:

python split_song.py

You can now use https://github.com/chavinlo/musicgen_trainer to finish building the model

Example:

cd musicgen_trainer
python run.py --dataset /full/path/to/your/data_model