Google Revolutionizes Translation with Gemini 3.5

A Major Breakthrough in Voice Translation with Gemini 3.5

Google has recently launched Gemini 3.5 Live Translate, an innovative audio model that enables near real-time translation of conversations in over 70 languages. This technology represents a significant advancement in the field of voice translation, providing a smoother and more natural user experience.

Since its inception twenty years ago, Google has made translation one of its primary fields of innovation in machine learning. Today, with Gemini 3.5 Live Translate, the company takes a new step by offering a model capable of translating a trillion words each month for billions of users worldwide.

Real-Time Voice Translations

The Gemini 3.5 model stands out for its ability to automatically detect over 70 languages and generate a voice translation that preserves the speaker's intonation and rhythm. Unlike traditional systems that wait for the end of a sentence to translate, Gemini 3.5 operates continuously, providing almost instantaneous translation. This approach reduces awkward pauses and maintains a fluid conversation, with only a few seconds of lag behind the speaker.

Deployment and Accessibility

Starting today, Gemini 3.5 Live Translate is available in several Google products. Developers can access it via the Gemini Live API and Google AI Studio in public preview, while businesses can experiment with it in private preview within Google Meet. Additionally, the Google Translate app on Android and iOS now incorporates this technology, making voice translation accessible to a broader audience.

A Robust and Adaptable Technology

Gemini 3.5 Live Translate is designed to process speech in real-time, even in noisy environments. This robustness allows the model to be used in various contexts, such as multilingual calls, meetings, classes, or live broadcasts. Thanks to the Gemini Live API, platforms like Agora, Fishjam, LiveKit, Pipecat, and Vision Agents can develop voice translation applications without worrying about the complex infrastructure required for real-time broadcasting.

Positive Feedback from Early Users

Companies such as Grab, CJ ENM, and LiveKit have already tested Gemini 3.5 Live Translate and expressed their satisfaction with the quality of translations, accuracy, and low latency of the model. Grab, for example, uses this technology to facilitate communication between drivers and passengers, handling over 10 million voice calls per month.

Upcoming Improvements for Google Meet

Google Meet will soon benefit from Gemini 3.5 Live Translate, expanding its voice translation offerings to over 70 languages, up from just five previously. This update will also enable conversations in more than 2000 language combinations, surpassing the current limitations of translation only from English. The Google Meet interface will also be updated to provide quicker access to voice translation.

Integration into Google Translate

The Google Translate app on Android and iOS now integrates Gemini 3.5 Live Translate, allowing users to enjoy smooth voice translation in over 70 languages. A new listening mode is also being rolled out for Android users, enabling them to hear translations directly through their phone's earpiece, without the need for external headphones.

Security and Accountability with SynthID

All audio generated by Gemini 3.5 Live Translate is marked with SynthID, an imperceptible watermark that ensures AI-generated content can be detected. This measure aims to prevent misinformation and ensure accountability in the use of this technology. For more information on Google's security practices, the model card is available for consultation.