Breaking

Google Unveils Gemini 3.5 Live Translate, a Speech Model Covering 70+ Languages

June 9, 2026 at 11:41 EDT

Foundation Models
Media Generation

On June 9, 2026, Google announced Gemini 3.5 Live Translate, its latest audio model delivering low-latency real-time speech translation. It supports more than 70 languages (2,000 language pairs) and processes speech as a stream rather than waiting for a speaker to finish, enabling near real-time speech-to-speech translation. Alongside the announcement, Google published an official blog post, a model card and API documentation, with availability through the Google Translate app, the Gemini Live API (public preview), Google AI Studio and Google Meet (enterprise private preview).

June 9, 2026 · Google

Gemini 3.5 Live Translate: speech that's translated while you're still talking

A single end-to-end audio model streams low-latency speech-to-speech translation, shifting from "translate after you finish" to "translate while listening" — preserving the speaker's tone, pacing and pitch.

70+

languages supported

2,000+

language pairs

model, audio in → audio out

The old chain vs. the new model

From three stitched-together stages to one end-to-end model.

Traditional turn-based pipeline

STT
recognition

→

Translate

→

TTS
synthesis

Gemini 3.5 streaming model

Audio in

→

Single end-to-end model
streams output mid-sentence

→

Audio out

What the single model does in real time

Multilingual input

Understands several languages mixed in one session — no config change.

Auto-detection

Identifies the spoken language and begins translating immediately.

Native audio

Generates natural speech, preserving intonation, pacing and pitch.

Noise robustness

Filters surrounding noise to stay clear in loud environments.

Gemini 3.5 Live Translate

Strength: naturalness, multilingual breadth
70+ languages / 2,000+ pairs
Single end-to-end model

OpenAI Realtime

Strength: low latency, conversational fluency
70+ input → ~13 output languages
Realtime API (audio-billed)

API pricing reference (estimated, per million units)

$3.50

input

$21.00

output

Available in public preview via the Gemini Live API and Google AI Studio; beta in the Google Translate app on Android and iOS, with Google Meet integration planned.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.