Good machine translation system is essential to facilitate language learning and content creation. In this paper, we compared the quality of machine-generated outputs of Sindhi Devanagri from various engines like AI4Bharat, Devanagri Translator and our own developed model. For an impartial evaluation, we had employed automatic evaluation metrics like BLEU, METEOR, TER, etc. The experimental results shown that our developed model has marked superior performance compared to the existing engines across all metrics. It gives more accurate and linguistically consistent outputs. These results validate our approach and its applicability to improve Sindhi Devanagri text generation.