Google Introduces Real-Time Speech Translation Model Supporting 70+ Languages

Google has unveiled Gemini 3.5 Live Translate, a new audio model designed for real-time speech-to-speech translation across more than 70 languages. The system automatically detects input languages and produces natural-sounding translated speech that preserves the speaker's original tone, intonation and pacing characteristics. Unlike traditional turn-by-turn translation systems, the model operates continuously throughout conversations, minimizing the awkward pauses between dialogue turns. The technology is rolling out across Google's ecosystem, starting with private preview access for Google Workspace business customers in Google Meet, while becoming available globally on the Google Translate mobile app for both iOS and Android devices. The model is already attracting attention from industry partners including ride-sharing platform Grab and developer platforms such as LiveKit and Pipecat, which are integrating the technology into their solutions. Security features include SynthID watermarking to identify AI-generated audio, and new capabilities like listening mode on Android, which streams translations directly through a phone's earpiece during calls.