Google launched its newest speech-to-speech translation system, Gemini 3.5 Live Translate, on Wednesday. The tool is built to support smoother multilingual conversations. The company states that the AI detects over 70 languages and produces translated speech while keeping the original speaker's tone, pacing, and intonation intact, delivering results with minimal delay. Gemini 3.5 Live Translate is now available in Google Translate, Google Meet, Google AI Studio, and the Gemini Live API for developers.

Gemini 3.5 Live Translate Handles Streaming Audio in Real Time

In its announcement, Google highlighted that this release continues its translation work that began nearly two decades ago with early machine learning efforts. The model targets multilingual meetings, live broadcasts, classes, customer support, and instant interpretation needs.

Google explained that Gemini 3.5 Live Translate manages live translation with low latency by processing audio streams as they arrive, outputting translated audio a few seconds after the speaker. It also removes the interruptions common in older turn-based systems.

The model automatically detects and switches among more than 70 languages without manual configuration. It stays reliable even in noisy or changing environments.

Google Meet is gaining upgraded translation tools that now cover more than 70 languages, a jump from the prior limit of five. Gemini 3.5 Live Translate is also reaching the Google Translate app on Android and iOS through the Live Translate feature, with output playable on compatible headphones.

On Android devices, a new Listening Mode lets translated audio play directly through the phone's earpiece.

To address synthetic media concerns, every audio clip from Gemini 3.5 Live Translate carries SynthID watermarking. The technology is embedded in the audio to help identify AI-generated material.

Developers can try Gemini 3.5 Live Translate in public preview via the Gemini Live API and Google AI Studio. Integrations with platforms such as Agora, Fishjam, LiveKit, Pipecat, and Vision Agents are intended to ease building and launching voice translation apps. The model is rolling out globally in supported Google products starting today.