Gemini 3.5 Live Translate: Google launches real-time voice translation for over 70 languages

Google has just taken another giant leap in the history of automatic translation. On June 9, 2026, the company announced Gemini 3.5 Live Translate, its latest audio model designed to perform speech-to-speech translation in almost real time, supporting over 70 languages. The technology promises to break down language barriers in a smooth, natural way without the awkward pauses that characterized earlier solutions.

Two decades of automatic translation reach a new level

The history of translation at Google began two decades ago as one of the company's first major machine learning experiments. Since then, the service has evolved impressively: today, over one trillion words are translated monthly for billions of users worldwide through the products of the Mountain View giant.

Gemini 3.5 Live Translate represents the next generation of this journey. More than just converting words from one language to another, the new model can capture and reproduce essential human elements of speech, such as intonation, rhythm, and tone, delivering an experience that sounds genuinely natural to anyone listening to the translation.

What makes Gemini 3.5 Live Translate different?

The major technical innovation behind Gemini 3.5 Live Translate lies in how the model processes speech. Unlike traditional turn-based translation systems — which wait for the speaker to finish an entire sentence before starting the translation — the new model generates the translated speech continuously, keeping pace with the speaker's rhythm in almost real time.

This balance is technically sophisticated: the system needs to constantly calibrate between waiting for enough context to ensure quality and translating immediately to maintain synchronization with the speaker. The result is fluid audio that is only a few seconds behind the original speech — a difference that is nearly imperceptible in natural conversations.

Among the most relevant technical differentials are:

• Automatic language detection: the model automatically identifies over 70 languages, without the need for manual configuration by the user
• Robustness to noise: the system is designed to work in noisy and unpredictable environments, such as live events, busy offices, or public spaces
• Preservation of vocal characteristics: intonation, cadence, and pitch of the original speaker are maintained in the translated voice
• Continuous flow processing: speech is processed as it arrives, allowing for a more fluid connection between people speaking different languages

Where is Gemini 3.5 Live Translate already available?

The rollout is being done gradually, reaching different audiences in distinct phases:

For developers

The model is already available in public preview through the Gemini Live API and the Google AI Studio. Developers interested in building real-time translation applications can start exploring the model's capabilities immediately, integrating it into their own products and services.

For businesses

Organizations using Google's corporate ecosystem will have access to Gemini 3.5 Live Translate in private preview on Google Meet, with a start expected in June 2026. This integration has transformative potential for international meetings, multilingual negotiations, and global collaboration among teams distributed around the world.

For the general public

Anyone can already experience the technology through Google Translate, available for both Android and iOS. The update arrives silently but powerfully to the app that is already a global reference in translation.

Impact for developers and companies

For software developers, Gemini 3.5 Live Translate opens up unprecedented possibilities. The ability to process speech in continuous flow means it’s possible to create:

• Multilingual customer service applications without friction
• Educational platforms with real-time classes for speakers of different languages
• Accessibility tools for minority language communities
• Telemedicine solutions that eliminate language barriers between doctor and patient
• Simultaneous translation systems for events, conferences, and live broadcasts

The robustness to noise is especially relevant for field use cases, where the environment is rarely controlled. This makes the technology viable in contexts such as journalism in conflict zones, humanitarian work in crisis areas, and tourism.

Google Translate: two decades of evolution culminating in this moment

It's hard not to be impressed by the trajectory that has brought Google to this point. In 2006, the company launched Google Translate as a pioneering experiment in machine learning applied to language. At the time, the technology was based on statistical models that, while revolutionary for the period, produced mechanical and often inaccurate translations.

With the advent of deep neural networks and, later, large-scale language models (LLMs), the quality of translations made exponential leaps. Today, with Gemini 3.5 Live Translate, the company is taking the next logical step: not only translating text, but translating the human voice in all its expressive richness.

What to expect next?

The launch of Gemini 3.5 Live Translate is undoubtedly just the beginning. As the model is refined based on feedback from developers and users, it is reasonable to expect:

• Expansion of the number of supported languages, focusing on less digitally represented languages
• Improvements in latency, further reducing the delay between original speech and translation
• Integration with other Google products, such as Google Meet for consumers, YouTube, and possibly Android at the operating system level
• Voice personalization, allowing translations to sound even closer to the original speaker's timbre

Conclusion: a new era for human communication

Gemini 3.5 Live Translate is not just a product update — it's a statement of intent about the future of human communication mediated by artificial intelligence. By democratizing high-quality simultaneous translation, Google is essentially making language barriers increasingly irrelevant for billions of people around the world.

If the promise holds true in everyday practice, we are witnessing one of the most significant contributions of generative AI to real life in recent years. Talking with anyone, in any language, naturally and fluidly — this ancient dream of humanity has never been closer to becoming an accessible reality for all.