Google has introduced Gemini 3.5 Live Translate, a new speech-to-speech AI model capable of delivering near real-time voice translation across more than 70 languages. The announcement marks Google’s latest effort to make multilingual communication feel as natural as speaking with someone in the same language.  

According to Google, Gemini 3.5 Live Translate automatically detects spoken languages and generates translated speech while preserving the speaker’s intonation, pacing, and pitch. The company says the model is designed to produce natural-sounding conversations rather than robotic translations.  

Key Features

Google highlighted several capabilities of the new model:

• Near real-time speech-to-speech translation

• Support for more than 70 languages

• Automatic language detection

• Preservation of voice characteristics and speaking style

• Low-latency audio processing

• Support for over 2,000 language combinations

Coming to Google Meet

One of the biggest integrations is with Google Meet. Speech translation in Meet will soon be powered by Gemini 3.5 Live Translate, expanding language support from just five languages to more than 70.

Google says users will be able to participate in multilingual meetings where conversations are translated in near real time, regardless of the languages being spoken. The feature is launching in private preview for select Google Workspace customers before a broader rollout later this year.  

Available in Gemini API and AI Studio

For developers, Google is making the model available through the Gemini API and Google AI Studio.

The new gemini-3.5-live-translate-preview model allows developers to stream audio in one language and receive translated audio output in another, enabling real-time multilingual applications, voice assistants, customer support systems, and communication tools.  

Early Enterprise Adoption

Google revealed that companies including Grab, LiveKit, Agora, CJ ENM, and Software Mansion have already been testing the technology.

Grab, for example, is exploring the model to help drivers and passengers communicate during pickups across different languages. Google says the platform handles more than 10 million voice calls each month, making translation accuracy and latency critical.  

Part of Google’s Agentic AI Strategy

The launch aligns with Google’s broader AI roadmap unveiled at Google I/O 2026. Gemini 3.5 Live Translate joins other recent releases including Gemini 3.5 Flash, Gemini Spark, Gemini Omni, and Antigravity 2.0 as part of Google’s push toward more capable, real-time, and agentic AI experiences.