deepl's new aure model captures tone and mood in real-time translation

deepl has launched its new aure model for real-time translation, which continuously analyzes both the translated text and the original audio as the conversation unfolds. this allows for the capture of tone, emotion, emphasis, and speaking pace from the original speaker's voice. previously, deepl voice generated translated speech from a static voice sample, limiting its ability to pick up on mood shifts or sarcasm. the aure model was trained using reinforcement learning to keep accents out of the output and supports english, german, spanish, french, italian, japanese, portuguese, and korean. deepl's update aims to improve the naturalness of its real-time translations.