Lyrics and piano music generation in Pytorch

Implementation of generative character-level and multi-pitch-level rnn models described in "Learning to generate lyrics and music with Recurrent Neural Networks" blog post.

The original jupyter notebook source of the blog can be found here.

Trained models can be downloaded via the following link. You can skip training and sample using provided models (follow jupyter notebooks below). Some examples of the piano song samples are available on youtube, and examples of lyris samples can be found in the original blog post.

Lyrics Generation

We are providing jupyter notebooks for training and sampling from generative RNN-based models trained on a song lyrics dataset which features most popular/recent artists. Separate notebooks are avalable for:

1. Training of the unconditional RNN-based generative model on the specified lyrics dataset (notebook file).
2. Sampling from a trained unconditional RNN-based generative model (notebook file).
3. Training of the conditional RNN-based generative model (notebook file).
4. Sampling from a trained conditional RNN-based generative model (notebook file).

Piano polyphonic midi song generation

We are providing jupyter notebooks for training and sampling from generative RNN-based models trained on a piano midi songs dataset. Separate notebooks are avalable for:

1. Training of the RNN-based generative model on the specified piano midi dataset (notebook file).
2. Sampling from a trained RNN-based generative model (notebook file).