Google Releases Gemini 3.5 Live Translate, a Streaming Speech-to-Speech Audio Model Covering 70+ Languages Across Meet, Translate, and the Live API

Google just announced Gemini 3.5 Live Translate . It is their latest audio model for live speech-to-speech translation. Speech-to-speech means spoken audio goes in, and translated spoken audio comes out. The model detects over 70 languages automatically and generates translated speech. It preserves the speaker’s intonation, pacing, and pitch in the output. Turn-by-turn systems wait for a speaker to finish before responding. Gemini 3.5 Live Translate generates speech continuously instead. It balances a trade-off between waiting for context and translating immediately. More context improves quality. Faster output keeps the translation in sync with the speaker. The result stays a few seconds behind the speaker throughout a session.

Gemini 3.5 Live Translate is a single audio model ( gemini-3.5-live-translate-preview ), not a chat assistant. It processes speech as the audio streams in, rather than after a full sentence. It handles multilingual inputs without manually configuring settings. Its noise robustness lets applications run in loud, unpredictable environments.

The model is rolling out across three surfaces. Developers get it in public preview through the Gemini Live API and Google AI Studio. Enterprises get a private preview in Google Meet starting this month. Everyone else gets it through the Google Translate app on Android and iOS.

The design difference matters for building real-time features. A conversational Live agent uses turn-based interactions. It relies on pauses, intent detection, and interruption handling. Live Translation uses continuous stream processing instead. It translates as the speaker talks, without waiting for turns to end.

To hold strict real-time latency thresholds, the translation path accepts audio input only. Text input is not supported in translation mode. The model also drops tool use and system instructions in this mode. That keeps it a focused translator pipeline rather than a general agent.

Developers configure translation inside the Live API session setup. You set a translationConfig block within the generationConfig . The targetLanguageCode field takes a BCP-47 code, such as "pl" or "es" . BCP-47 is the standard format for language tags like en or pt-BR . It defaults to "en" . The echoTargetLanguage boolean controls input that is already in the target language. When true , the model echoes that speech. When false , it stays silent. You can also enable inputAudioTranscription and outputAudioTranscription for text transcripts.

Audio formats are fixed. Input is raw 16-bit PCM at 16kHz, mono, little-endian. Output is raw 16-bit PCM at 24kHz, mono, little-endian. PCM is uncompressed raw audio. You send audio in chunks of 100ms. For client-side apps, ephemeral tokens on the v1alpha endpoint avoid exposing your API key.