Poetry translation is one of the most challenging forms of literary translation. It not only involves language translation but also deals with the sense, emotion, and message of the poem. It also involves knowledge of the culture from which the poem has been taken. In this paper, we present a study of the translation of Hindi poetry into English with the help of various models like mBART, seq2seq, Neural Machine Translation (NMT) and a hybrid of GPT and LSTM. We have used Hindi poetry as our dataset and carried out the evaluation based on linguistic aspects, fluency and keeping the poetic sense intact. We have used different metrics to evaluate the models qualitatively and quantitatively including BLEU scores and human evaluation. The study shows the main challenges posed by Hindi poetry translation, that is the literal translation of a phrase, the inability of a model to recognize a phrase that uses metaphors, the absence of a word in the training corpus that makes the model wrongly translate it, the inability to recognize the adjective in a noun-adjective pair, a model’s inability to recognize a comparative degree adjective, the translation of a literal word, and the inability to recognize plural nouns in Hindi.