DeepL unveils Aure: voice translation aims to follow a speaker’s tone
Translating words is not the same as conveying how they are spoken. On 11 November 2026, DeepL introduced Aure, a new model for DeepL Voice that the company says adapts translated speech in real time to changes in tone, emotion, emphasis and pace.[1] RuntimeWire’s accessible opening paragraph also reports the launch and attributes that capability to the vendor; it is not an independent performance test.[2]
## From a recognisable voice to a changing voice
DeepL describes a system that continuously analyses source audio alongside translated words to shape how a sentence is delivered in another language.[1] The claimed advance concerns expression throughout a conversation, rather than voice identity alone.[1] That technical promise should not be confused with certainty that every nuance will be reproduced correctly.
## Where the company says it is available
The company says the new capabilities apply automatically, without additional configuration, across DeepL Voice for Online Meetings, DeepL Voice for In-Person Conversations and DeepL API for Voice.[1] Its initial language list is English, German, Spanish, French, Italian, Japanese, Portuguese and Korean; other languages remain supported through existing voice models.[1] This announcement concerns DeepL Voice: it does not by itself establish free access or inclusion in every translator subscription.
## What to check before relying on it
For an international meeting, natural-sounding speech is only part of the question. Accuracy, delay, overlapping speakers and preservation of intent all deserve testing: a plausible tone does not by itself guarantee a faithful translation. These are editorial evaluation criteria, not findings from a test conducted by this publication.
DeepL also says the service does not permanently store meeting transcripts or translations and deletes this data from memory when a session ends.[1] That is the provider’s statement, not an independent data-handling audit. The verified news is the model announcement and its stated scope; real-world effectiveness remains to be measured.
## Sources
[1] https://www.deepl.com/sv/press-release/aure-voice-ai-model — DeepL unveils Aure — comunicato ufficiale 11 novembre 2026
[2] https://runtimewire.com/article/deepl-aure-voice-model-tone-translation — RuntimeWire — DeepL launches Aure (solo incipit accessibile)