Make seed lyrics. Feed it to the neural network to generate one character after the seed lyrics, ‘b’. I took input char length 100.
input: 'I want to ' -> output 'b'
Append new character to the seed lyrics and remove the very first character.
new input : ' want to b'
Feed the new seed lyrics into the neural network and iterate the above process as many as you want.
'I want to ' -> ' want to b' -> 'want to be' -> .... -> 'ing of pop'
In the end, you might get something like ‘I want to be king of pop’
This process is known as teacher forcing: training neural network that uses model output from a prior time step as an input.
I trained model for 15 Epoch for batch size 32 , for more accuracy increase the epoch for training (400, 600 etc.)
Lstm network architect : https://colah.github.io/posts/2015-08-Understanding-LSTMs/
https://levelup.gitconnected.com/lyrics-generation-using-lstm-5a5a0bcac4fa
https://medium.com/@shivajbd/understanding-input-and-output-shape-in-lstm-keras-c501ee95c65e