• New Voice AI models and product updates deliver voice preservation that carries more of each speakers' tone, emotion and expression; and real-time translation with low latency
• DeepL also unveils new desktop app for seamless multilingual online meetings, with voice to voice capabilities now generally available in 30+ languages
• DeepL Voice is scaling fast, expanding across enterprise use-cases and live experiences including powering real-time translation across all 50 stages at Salesforce Dreamforce 2026

COLOGNE, Germany, Sept. 15, 2026 /PRNewswire/ -- DeepL, a global leader in Language AI, today announced major upgrades to its Voice AI technology. DeepL Voice has consistently led in translation accuracy, ranked #1 in Slator's independent assessment of AI-translated captions. Now, new Voice models take the next step, delivering voice preserving that retains not just what people say, but how they say it. Users of Zoom, Microsoft Teams and Google Meet can now talk in real time across languages with DeepL Voice, keeping their voice, tone and pace, making multilingual conversations feel far more natural.

"Voice is how people show up, and it carries far more than words," said Jarek Kutylowski, CEO and Founder of DeepL. "With these latest updates to DeepL Voice, we're moving beyond just subtitles to audio-native systems that preserve identity, emotion, and timing in real time. We've built models that make you forget you're speaking through AI, making voice technology feel native, not processed. These models bring the human layer into real-time multilingual conversations."

Alongside this, DeepL has launched a new desktop app for Voice that delivers a unified translation experience across meeting platforms, with no extra integrations needed. Voice to voice translation is now generally available for online meetings, in-person conversations and via the DeepL API.

What's new in DeepL Voice

Real-time voice translation has made significant progress on speed and accuracy. The next frontier is preserving the distinct qualities that make one person sound different from another – the subtle vocal cues that deeply shape how spoken language is understood. With its new models, DeepL Voice carries more of that human layer across languages, preserving tone, rhythm, pacing and intonation so cross-language conversation sounds less synthetic and more like the people behind the words. DeepL's latest Voice AI models deliver:

• Voice preservation: Each person's voice remains distinct and recognizable across 14 supported languages, even when multiple participants are being translated simultaneously
• Natural expression: Questions sound like questions; hesitation, urgency or excitement carry through the translation
• Low latency: The models maintain real-time pace while preserving far more contextual and human information from speech
• Expanded language support: Voice to voice translation now available across 30+ languages

"To get this right, we had to solve several hard technical challenges at once, preserving voice identity, while maintaining translation quality and low latency," said Sebastian Enderlein, CTO, DeepL. "This is a significant technical step forward in what DeepL Voice can do, and sets the bar high for what people should expect when communicating across languages. It also opens up entirely new opportunities for businesses - from how they operate and serve customers to the multilingual products and experiences they can build, and other use cases we haven't even imagined yet."

DeepL also unveils its new Voice desktop app, available for Windows and Mac, providing one consistent translation experience across online meetings.

The DeepL Voice app works alongside major meeting platforms  Zoom, Microsoft Teams and Google Meet, automatically detecting calls and providing an always-on overlay for live translated subtitles. Users can follow captions while presenting or taking notes; switch to voice to voice translation when they'd rather listen than read; and adjust caption size, background audio levels and translation settings. The app also enables users to access new Voice capabilities as they roll out, without being dependent on platform-specific updates.

Scaling real-time voice translation across work and live events

These updates come as demand for DeepL Voice grows across industries and use cases. Enterprises across manufacturing, retail, life sciences, professional services and more use DeepL Voice to power multilingual collaboration across global teams, customers, and partners; enabling them to work more seamlessly across languages, improve productivity, engage diverse workforces and enter new markets faster. Through the DeepL Voice API, developers are also embedding real-time voice AI into contact centers, meeting tools, frontline apps and custom products.

That demand also extends to large-scale events and experiences. DeepL is a Voice AI partner for Salesforce Dreamforce 2026 (September 15–17, San Francisco), powering real-time translation across 50 stages and over 1,000 sessions. The company is also showcasing its latest Voice updates, including voice to voice translation, voice preservation and enhanced expression, at the DeepL demo booth onsite.

Learn more and try DeepL Voice at https://dee.pl/r2ett.

About DeepL

DeepL is a global AI company building the language infrastructure that powers global business. More than 200,000 business teams and millions of individuals use DeepL's Language AI platform to communicate globally, collaborate and operate across languages in real time. By combining breakthrough AI models with enterprise-grade security and privacy, DeepL enables organizations to work seamlessly across markets and cultures. Founded in 2017 by CEO Jarek Kutylowski, DeepL now has more than 900 employees and is backed by leading investors including Benchmark, IVP and Index Ventures. Learn more at www.deepl.com.