DeepL Aure Brings Tone Into Live Translation

Key takeaways

• DeepL announced Aure on November 11, 2026, a new voice AI model designed to carry tone, emotion, emphasis and speaking pace into real-time translations.

• The model is being applied across DeepL Voice for Online Meetings, In-Person Conversations and the DeepL Voice API.

• Initial language support covers eight languages, including English, Spanish, Japanese and Korean; the company has not announced a date for additional languages.

DeepL says its new Aure model can make live speech translation reflect not only the words a person says but also how they say them. Announced on November 11, 2026, Aure continuously analyzes the original audio alongside translated words to adapt delivery as a conversation changes, with initial support for eight languages.

What DeepL announced on November 11

DeepL introduced Aure as the new voice model powering its real-time translation product, DeepL Voice. The company says the model adjusts translated speech to reflect changes in a speaker's tone, emotion, emphasis and pace during a conversation.

That is a change from the approach DeepL describes in its earlier voice technology. Previously, the system used a sample from the beginning of a conversation to shape the translated voice. That initial sample could preserve a speaker's recognizable voice, but it could not fully follow later shifts in mood, emphasis or speed.

Aure is designed to keep analyzing the incoming audio as the conversation continues. DeepL says it uses both the translated words and the original audio to shape how the translated sentence is delivered. The company says this can help convey expressive differences such as sarcasm or a change from a calm tone to a more urgent one.

The launch announcement says Aure's capabilities are applied automatically across three DeepL products: Voice for Online Meetings, Voice for In-Person Conversations and the DeepL API for Voice. The initial supported languages are English, German, Spanish, French, Italian, Japanese, Portuguese and Korean. DeepL says more languages will follow, but it has not provided a timetable.

Why tone-aware translation could matter

Analysis: Most people judge a conversation by more than its literal words. A question can sound curious, impatient or skeptical depending on the speaker's delivery. A negotiation can become more urgent, and a colleague's reassurance can sound less convincing if the translated voice misses the original emphasis.

If Aure performs as DeepL describes, carrying those changes into another language could make remote meetings and in-person conversations feel less mechanical. It may be useful for international teams, customer support, travel and business discussions where tone can change how a message is understood.

The potential benefit is not that a translation becomes infallible. The model still has to interpret speech in noisy rooms, distinguish the speaker from background sounds and transfer expression between languages with different rhythms and pronunciation. A fluent-sounding voice also does not guarantee that the underlying translation is correct.

DeepL says it addressed some of those challenges by separating speaker identity, prosody and spoken content during training. Prosody means the rhythm and expressive features of speech, including intonation, emphasis and pace. The company says it also uses reinforcement learning to refine speech quality and translation accuracy.

Independent coverage by The Decoder and AI Affairs describes the same shift from a fixed voice sample to continuous analysis of the speaker's audio. Those reports explain the product change; they are not independent benchmark tests proving that Aure outperforms competing systems.

What DeepL Voice users should do now

If your organization already uses DeepL Voice, the first step is to check which product and language combination your team relies on.

• For online meetings: confirm that your account uses DeepL Voice for Online Meetings and check the product's current language availability.

• For in-person conversations: review whether your organization's DeepL Voice setup supports the languages your staff and customers use.

• For developers: consult the DeepL Voice API information and current developer documentation before changing an integration.

• For important conversations: keep a human review step for contracts, medical discussions, legal matters and other exchanges where a mistranslated phrase could have serious consequences.

The announcement does not provide a separate price for Aure or describe a new paid add-on. DeepL says the capabilities are applied automatically across the named Voice products, so users should check their existing plan and product settings rather than assume that a new subscription is required.

What is still unknown and what to watch next

DeepL has not published a detailed independent evaluation of Aure's accuracy across accents, noisy environments, specialist vocabulary or emotionally complex conversations. It also has not given a rollout schedule for languages beyond the initial eight.

The company's description sets out the intended behavior, but real-world performance will depend on the speaker, language pair, microphone quality, background noise and subject matter. Users should look for independent testing that measures both translation accuracy and whether changes in tone are carried over correctly.

DeepL also says meeting transcripts and translations are not permanently stored and are deleted from memory when a session ends. Organizations handling confidential conversations should still review the product's current data-processing terms and their own privacy requirements before enabling voice translation.

What is DeepL Aure?

Aure is a voice AI model from DeepL that powers expressive, real-time speech translation. It is designed to adapt translated speech to changes in a speaker's tone, emotion, emphasis and pace.

Which languages does Aure support at launch?

DeepL lists English, German, Spanish, French, Italian, Japanese, Portuguese and Korean. The company says additional languages are planned but has not announced dates.

Does Aure translate sarcasm and emotion perfectly?

DeepL says Aure is designed to carry expressive cues, including sarcasm, into translated speech. The company has not published independent benchmark results showing how reliably it handles sarcasm or emotion across different speakers and settings.

Where will users find Aure?

DeepL says the model's capabilities are applied across DeepL Voice for Online Meetings, DeepL Voice for In-Person Conversations and the DeepL API for Voice. Users should check their product documentation and account setup for the available features.

Does Aure store meeting transcripts?

DeepL says it does not permanently store meeting transcripts or translations and deletes that data from memory when a session ends. Organizations should consult DeepL's current privacy and data-processing terms for details relevant to their own use.

Bottom line

Aure targets a real limitation in speech translation: preserving the speaker's changing delivery, not just their words or voice identity. DeepL's announcement is a meaningful product update, but users should treat its performance claims as company claims until independent testing is available and check the supported language list before relying on it for important conversations.

Sources: DeepL's Aure announcement, November 11, 2026 · The Decoder's independent coverage · AI Affairs coverage