Google Unveils Gemini 3.5 Live Translate for Real-Time Multilingual Conversations

A new audio model from Google aims to make real-time multilingual conversations feel more natural, allowing people to communicate freely without language barriers. The technology, called Gemini 3.5 Live Translate, continuously listens, translates and speaks in a matter of seconds, mimicking the flow of natural speech patterns.

The model supports over 70 languages, enabling thousands of language pairings within the same conversation. It automatically detects spoken languages and can operate in noisy environments, handling background sounds, overlapping voices and informal speech patterns with ease.

Unlike traditional translation systems that process speech in turns, Gemini 3.5 Live Translate performs continuous streaming translation, reducing awkward pauses, interruptions and delays. This makes it suitable for a wide range of use cases, including customer support calls, guided tours, classrooms, ride-sharing services and live broadcasts.

The system also prioritizes speech quality, attempting to preserve elements of the original speaker's delivery, such as pacing, intonation and emotional tone. This helps translated speech sound more natural and makes conversations easier to follow.

Google is making Gemini 3.5 Live Translate available to developers and partners, who can integrate it into meetings, communication platforms and mobile applications. The technology has the potential to move live translation beyond occasional demonstrations and into everyday communication, making cross-language interactions more practical for businesses, organizations and individuals alike.