Music Generation Using RNN-LSTM with GRU

Recurrent neural networks (RNNs) and long short-term memory (LSTM) networks are two methods for music production that have been studied by researchers. A lot of LSTM networks have been employed to create character-level musical scores. However, in order to produce pleasing and grammatically accurate sheet music, these LSTM models necessitate a significant amount of training time. As an alternative to LSTM models in this study, we use Gated Recurrent Unit (GRU) networks, which have three gates and do not maintain an internal cell state. Peer evaluations of the resulting music's quality use qualitative criteria. The suggested GRU-LSTM model is compared subjectively with another model without similar techniques.