The difference of application of RNN on NLP and acoustic features training

Question

For natural language processing. If we are trying to do a text generator.

For example.

LSTM(128,input_shape(60,50))

60 means the input