The term music comes from the Greek translation “mousike” that stands for “the art of muses”. In layman terms, it is a sequential string of sounds that people like for the purpose of entertainment, rejuvenation etc. The intriguing part of this research lies in the ability of a machine to manipulate music. The aim is to use a part of given composition (MIDI File) to generate further music via extrapolation from Long short-term memory (LSTM) which is a which is a neural network used in deep learning an d artificial intelligence, then increasing the efficiency of the model using the relation between training loss and number of epochs and finally comparing our LSTM model to traditional RNN model.