DeepL launches Aure, a more expressive voice-translation model

DeepL, the language-AI company, introduced a new voice-translation model called Aure for DeepL Voice, its real-time speech-translation product. The company describes Aure as its most dynamic and lifelike Voice AI model yet. DeepL says its earlier approach generated an expressive translated voice from a single sample taken early in a conversation, capturing who was speaking but not how — so it missed later shifts in emotion or pace, or a delivery that cut against the words themselves, like sarcasm. Aure instead analyzes speech continuously, adjusting the translated voice as a speaker’s tone, emotion, emphasis and pace change throughout the conversation.

CEO and founder Jarek Kutylowski said the goal is for translated speech to carry more of a speaker’s personality across languages: “With Aure, we can capture even the subtlest shifts in expression as they happen and carry them into another language, in someone’s own voice and in real time.” Aure builds on a voice-preservation feature DeepL launched previously, which let people sound like themselves rather than a generic synthetic voice; Aure adds the ability to track how that voice shifts throughout a conversation rather than fixing it from an opening sample.

To do this, Aure draws on both the translated words and the original audio to shape delivery, while stripping out background noise, echo and signal distortion and isolating a speaker’s voice identity and prosody. Because DeepL Voice keeps no permanent record of meeting transcripts or translations — it clears that data once a session ends — Aure has little recorded history of any one speaker to learn from, so it adapts in real time, before a sentence even finishes. DeepL says it combined several training methods so a speaker’s identity, the way they speak, and what they say can each move into a new language on their own, then applied reinforcement learning to keep translation accuracy and speech quality intact while avoiding an exaggerated accent.

Aure is live now, with no extra setup, across DeepL Voice for Online Meetings, DeepL Voice for In-Person Conversations and the DeepL API for Voice. It initially covers eight languages — English, French, German, Italian, Japanese, Korean, Portuguese and Spanish — with more to follow; other languages continue to run on DeepL’s existing voice models. DeepL says customers including Pioneer Corporation, Aramark, Avendra International and Inetum are already using DeepL Voice, and that the product was recently named to Fast Company’s 2026 Next Big Things in Tech list in the Applied AI category.

Why it matters

For global businesses running meetings and negotiations across languages, tone often carries as much meaning as the words themselves — sarcasm, urgency or enthusiasm that a flat, single-snapshot voice translation could miss. If Aure performs as DeepL describes, that narrows a gap rival voice-translation and interpretation tools still have to close.