RNN-based generative models for speech. - View it on GitHub