Listen and compare speech outputs from different S2ST models across various emotion categories. Each row shows the same source utterance translated by each model.