Google previews continuous conversation with Gemini Live Translate

Google announced Gemini 3.5 Live Translate on June 9 as a system for continuous voice translation. The company said it automatically detects more than 70 languages and can cover more than 2,000 language combinations in one meeting. The product is designed to keep a conversation moving without requiring each person to select a language and wait through a rigid sequence of separate turns.

Google said the model aims to preserve intonation, pacing and pitch rather than reducing speech to words alone. Those qualities can carry meaning in a conversation, particularly when speakers interrupt, hesitate or emphasize a phrase. The announcement describes the intended behavior, but it does not establish equal accuracy or naturalness across every supported language pair.

A public preview opened through the Gemini Live API and Google AI Studio. Integrations with Google Translate and Google Meet were announced as planned additions, which means availability in the preview should not be confused with a completed rollout across Google's consumer and workplace products. Pricing, regional access and model identifiers should be checked in current documentation before implementation.

The shift is useful for travel, meetings, customer support and language learning because the translation layer can remain active while people talk. Its limits are also easier to expose in those settings. Accents, mixed-language speech, overlapping voices and background noise can all affect a system that performs well in a clean demonstration.

Language count is therefore only the opening specification. A meaningful comparison will need the same recordings, speakers and scoring method across systems, with native speakers assessing whether tone and meaning survive. Until that evidence exists, Google's reach and quality statements should remain company claims rather than a ranked verdict on the best translator.