YOU GUYS! Google just announced Gemini 3.5 Live Translate, which is a MASSIVE leap over the robot-voice translations we've been dealing with for years and you need to read this because it actually sounds like a person talking — not some flat synthesizer. It's a speech-to-speech model that translates between more than 70 languages while matching the original speaker’s intonation, pacing, and pitch in near real time (seconds behind). The Gemini 3.5 family already rolled out Flash earlier this year and Google is expecting a Pro version to drop shortly as well. For developers using the Gemini Live API or AI Studio, it processes audio continuously with no manual configuration needed — automatic language detection plus an actual background noise filter for messy environments — which means less friction and more scale from day one.
The rollout schedule matters too: enterprise customers get this in Google Meet first starting today before a wider release, AND it's coming to the Translate app on both Android and iOS soon! The best part? You don't need Pixel Buds anymore because any Bluetooth headphones will work — or if you have no headset at all, Android users can use "listening mode" by holding their phone up like they’re on a call. There is one catch though: every output gets SynthID watermarks baked into the waveform so it's permanently tagged as AI-generated and can't be removed. Still, hearing someone speak Spanish and get spoken English back through your earpiece while keeping your own voice characteristics intact — that's genuinely a decade of technology condensed into one model!
Source: https://arstechnica.com/ai/2026/06/google-announces-gemini-3-5-live-translate-for-instant-voice-to-voice-translation
The rollout schedule matters too: enterprise customers get this in Google Meet first starting today before a wider release, AND it's coming to the Translate app on both Android and iOS soon! The best part? You don't need Pixel Buds anymore because any Bluetooth headphones will work — or if you have no headset at all, Android users can use "listening mode" by holding their phone up like they’re on a call. There is one catch though: every output gets SynthID watermarks baked into the waveform so it's permanently tagged as AI-generated and can't be removed. Still, hearing someone speak Spanish and get spoken English back through your earpiece while keeping your own voice characteristics intact — that's genuinely a decade of technology condensed into one model!
Source: https://arstechnica.com/ai/2026/06/google-announces-gemini-3-5-live-translate-for-instant-voice-to-voice-translation