Google ships Gemini 3.5 Live Translate across consumer, enterprise and developer tools
The audio model streams speech-to-speech translation across 70-plus languages, with Meet access limited to private preview this month.
By Ryan Merket · Published
Why it matters
Live translation is becoming a distribution fight, not just a model benchmark. Google can push Gemini into Translate, Meet and developer tooling at once, turning a research capability into a platform feature before smaller voice AI startups can win the same surface area.
Google released Gemini 3.5 Live Translate on Tuesday, putting its latest audio model for near real-time speech-to-speech translation into Google AI Studio, Google Translate and, in limited form, Google Meet, according to a company blog post.
The product is being fronted by Anuda Weerasinghe, a Google product manager, and Tony Lu, a senior staff software engineer. Google says the model automatically detects more than 70 languages and produces translated speech that preserves a speaker's intonation, pacing and pitch.
The technical claim Google is making is not just broader language coverage. The company says Gemini 3.5 Live Translate processes streamed speech continuously rather than waiting for a speaker to finish, keeping translated audio a few seconds behind the original speaker while trading off immediacy against context. That is the central product bet: translation that feels closer to a live conversation than a turn-by-turn interpreter.
Distribution is the other half of the announcement. Developers get public preview access through the Gemini Live API and Google AI Studio; enterprises get a private preview in Google Meet starting this month; consumers get access through Google Translate on Android and iOS. Google also says Translate handles more than a trillion words per month across its products, a scale that gives the model an immediate testing surface its startup rivals cannot match.
The unanswered question is performance outside demos: noisy rooms, overlapping speakers, accents and high-stakes enterprise meetings are where live translation usually breaks. Google says the model is built for noise robustness and multilingual inputs without manual configuration, but the rollout still starts with previews for developers and Meet customers.