Twenty years ago Google translation began as one of our pioneering machine learning experiments, aiming to transform the science of language into the magic of human relationships. This experiment has come a long way – over a trillion words are translated every month for billions of users of our products.
Today we’re taking the next step by releasing Gemini 3.5 Live Translate, our newest audio model for live speech-to-speech translation.
The model automatically detects over 70 languages and generates fluent, natural-sounding translated speech that preserves the intonation, tempo and pitch of the speaker’s voice. Unlike step-by-step systems that wait for the speaker to finish speaking before responding, 3.5 Live Translate generates speech continuously, balancing the trade-off between waiting for the quality of the context to improve and translating immediately to stay in sync with the speaker. It delivers velvety audio without any awkward pauses and stays just a few seconds behind the speaker throughout your session.
Gemini Live Translator 3.5 will be available starting today on Google services:
Create with Live Translation 3.5
Gemini 3.5 Live Translate processes speech as it streams, enabling a more seamless connection between languages. The model supports multilingual data entry without the need to manually configure settings. At the same time, its resistance to interference ensures that applications can cope with clamorous, unpredictable environments. You can exploit its capabilities to facilitate live translation of multilingual conversations, meetings, lessons, broadcasts and more.