aykutakar / DeepRemix

Remixing Music with Deep Learning

DeepRemix

Remixing Music with Deep Learning

An interface in which you input a number of songs by one artist, and a song by another artist. Then a LSTM neural network in a GAN framework will output a remix of the song in the style of the first artist. Inputs are being parsed as Mel-Frequency Cepstrum Coefficients. Testing is done using PyTest and TravisCI.

For more information see DeepRemix.com.

Dependencies

• Python 2.7
• Keras
• Scipy
• Numpy
• librosa
• PyTest

Usage

Clone the repository, copy the song you'd like to remix into ./data/to_remix and run

python music_parsing.py

Once the song has been put in an appropriate representational state (MFCC), you can now run it through the neural network by running:

python scriptthatrunsnetworkgoeshere.py

This will save the output song into ./data/output_wavs

Travis CI

To test a build, do the following:

• Commit and push something to the repository...