Generate music snippets using Recurrent Neural Networks and display them on the web.
Neural Noise is a series of Python scripts that interact with external programs in order to:
- Generate a training set of music represented as character data, using ABC notation or some variation thereof;
- Train a Recurrent Neural Network (RNN) with the data, using the excellent char-rnn package;
- Generate new music snippets using checkpoints output by char-rnn;
- Finally, allow the user to randomly browse the new snippets using a web interface.
This is a complicated mish-mash (hack) of software, and getting it running is not for the faint of heart. Still, in this README we will attempt to document the complete set of steps that were used in installing char-rnn, massaging the training data, running the training, generating the snippets, and finally getting the web interface up and running.