- Google has officially announced “Gemini 3.5 Live Translate,” a new speech model rolled out as part of its latest “Gemini 3.5” AI model series.
- This AI speech model automatically detects over 70 languages and generates smooth, natural translated audio while maintaining the speaker’s original intonation, pace, and pitch.
- It will be initially available through Google Translate and Google Meet.
On Tuesday, June 9, 2026 (local time), Google officially announced “Gemini 3.5 Live Translate,” a state-of-the-art speech model launched as a key component of the new “Gemini 3.5” AI model series.
The “Gemini 3.5 Live Translate” speech model is designed to automatically detect over 70 languages, generating smooth and natural translations while preserving the speaker’s original intonation, pace, and pitch. Unlike “turn-by-turn” systems that wait for a speaker to finish before responding, this model is characterized by its ability to generate translated audio continuously.
Furthermore, “Gemini 3.5 Live Translate” supports multilingual input, serving as an AI speech model that enables real-time translation between multiple languages.
By balancing context awareness with the speaker’s natural flow, the system provides smooth audio without awkward pauses. As a result, the latency between the speaker and the translated output is expected to be only a matter of a few seconds throughout an entire session.
Additionally, “Gemini 3.5 Live Translate” offers excellent noise resistance, ensuring smooth performance even in noisy or unpredictable environments. This makes it highly effective as a live translation system for multilingual calls, meetings, lessons, and live broadcasts.
“Gemini 3.5 Live Translate” began rolling out across Google products on Tuesday, June 9, 2026, starting with Google Translate and Google Meet.
Source: Google (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-live-3-5-translate/)