The purpose of this project is to train a Recurrent Neural Network to reproduce one Canto of the Divina Commedia, written by Dante. We use Tensorflow 2 to build an RNN capable of understanding the structural complexity of this Poem. The network needs to learn to write Terzine composed of three rows, each row needs to have eleven syllables, moreover following the ABABCB rhyme structure.