DeepL launches Aure speech translation model with real-time tone and emotion

On November 11, 2026 (local time), DeepL launched Aure, a voice AI translation model that the company describes as its most dynamic and realistic voice model to date. Serving as the underlying engine for DeepL Voice, Aure continuously adapts translated speech in real time, reproducing the speaker's intonation swings, emotional shifts, emphasis, and pacing. According to DeepL, this allows voice translation to go beyond conveying literal words and sentences, preserving the personal traits, subtle emotions, and expressive richness that give human conversation its naturalness and impact. DeepL says Aure builds on its recently introduced voice preservation technology, which already lets people speak across languages in their own highly recognizable vocal timbre rather than a uniform mechanical synthetic voice. If that breakthrough retained a speaker's vocal characteristics, Aure goes further: it captures the dynamic movement and immediate reactions of a voice throughout a conversation, injecting more spontaneity and authentic character from natural human speech into real-time translation. Aure is now available as the powering core of DeepL Voice for web conferencing, DeepL Voice for face-to-face conversations, and the DeepL Speech API. The model initially supports eight languages: English, German, Spanish, French, Italian, Japanese, Portuguese, and Korean, with further languages to be added gradually.