Aure

Aure is the Voice AI translation model DeepL unveiled on 11 November 2026. Officially, it is DeepL’s most natural and lifelike voice translation model yet, and it now powers DeepL Voice. The product page is Introducing Aure. The press release is on PR Newswire and is listed on DeepL’s press page. The official account @DeepLcom posted at 2026-11-11T08:15:00Z. In the release, founder and CEO Jarek Kutylowski said Aure is meant to carry even subtle shifts in expression into another language, in the speaker’s own voice and in real time.

Officially, DeepL had already launched voice preservation, so a cross-language conversation could keep a recognizable personal voice instead of a generic synthetic one. Aure is the next step: it is not only “whose voice this is,” but “how they are speaking right now.” Previously, DeepL Voice largely decided the translated voice from a static sample at the start of a conversation, which could not follow a change in emotion, emphasis or pace, or delivery that contradicts the words, such as sarcasm. Aure uses both the translated words and the source audio, and starts adapting before a sentence is finished. DeepL says that means isolating the speaker’s voice identity and prosody in noisy settings, filtering background noise and echo, and carrying a voice across languages without exaggerating an accent or distorting how someone sounds. The company also says it does not permanently store meeting transcripts or translations, and deletes this data from memory when a session ends, so Aure has to adapt to each speaker in real time without a long-term cross-language sample.

On the method, DeepL says it combined training techniques that separate voice identity, prosody and spoken content, then used reinforcement learning so translation accuracy and speech quality stay intact while the translated voice sounds natural and free of accent. The product page says Aure matches a voice dynamically during the live session, without creating or storing a permanent voice profile; data is encrypted in transit and is not used to train DeepL’s models. DeepL Voice also filters background noise, other speakers and echoes so it can work in open offices or on a factory floor. The page states that, in online meetings, DeepL Voice has 76% fewer critical translation errors than platform-native translation tools; that figure is the company’s, and we have not independently reproduced it.

No extra setup is required. Aure is applied automatically across DeepL Voice for Online Meetings (Microsoft Teams, Zoom and Google Meet), DeepL Voice for In-Person Conversations, and the DeepL API for Voice. It initially supports eight languages — English, German, Spanish, French, Italian, Japanese, Portuguese and Korean — with more to follow. For languages Aure does not yet cover, DeepL Voice automatically uses the previous voice models. The release names Pioneer Corporation, Aramark and Avendra International, and Inetum as organizations already using DeepL Voice, and notes that DeepL Voice was listed in Fast Company’s 2026 Next Big Things in Tech awards in the Applied AI category.

Key links:

• Product: https://www.deepl.com/products/voice/aure ; press release: https://www.prnewswire.com/news-releases/deepl-unveils-aure-its-most-lifelike-voice-ai-translation-model-yet-302935694.html ;
• Official post: https://x.com/DeepLcom/status/2120430075149443191 ; DeepL Voice: https://www.deepl.com/en/products/voice ;
• This is a specific DeepL speech-translation model name. Do not treat it as the DeepL Voice product itself, or as the same news as the September voice-preservation update.

Why it matters

Real-time voice translation could already keep a recognizable personal voice; Aure is a named model whose point is to follow tone and rhythm as they change, not a vague feature tweak. It is applied automatically across DeepL Voice’s meeting, in-person and API paths, so teams already on DeepL Voice have little to migrate. On security and data, the official pages say there is no permanent voice profile or meeting transcript, and the data is not used for training — an important claim for meeting use.

Language coverage is still a limit: Aure starts with eight languages, and other languages stay on the previous models. The 76% reduction in critical errors is a company figure and should be cited as such. This entry is not related to Aurelio Finance, which is already in the database.

Limits

This page is based on the official product page, the press release and the @DeepLcom announcement. It does not include unverified listening tests. Plans, seats and API pricing should be read from DeepL’s pricing pages at the time of use.

1. Read the product page https://www.deepl.com/products/voice/aure and the press release https://www.prnewswire.com/news-releases/deepl-unveils-aure-its-most-lifelike-voice-ai-translation-model-yet-302935694.html , and cross-check the announcement at https://x.com/DeepLcom/status/2120430075149443191 ;
2. For the eight supported languages, use DeepL Voice in online meetings, in-person conversations or the Voice API — Aure is applied automatically; other languages fall back to the previous Voice models ;
3. When citing “most natural Voice model,” the eight-language list, the lack of a stored voice profile, or the 76% reduction in critical translation errors, say the source is the 11 November 2026 official page or press release, and re-check that source ;
4. Do not treat Aure as the entire DeepL Voice product, and do not confuse it with Aurelio Finance; this page only restates the official product page, press release and announcement — re-check the upstream source before citing it.