To beat Apple, Google skips the earbuds

The most revealing feature in Google's new translation model is also the most mundane. On Android, a setting called "listening mode" lets a user hold the phone to the ear, as if taking an ordinary call, and hear a foreign conversation arrive in their own language through the earpiece — no headphones, no second gadget, nothing to buy. Google unveiled the model, Gemini 3.5 Live Translate, this week, wrapped in the kind of specifications that earn a demo reel: automatic detection of more than 70 languages, speech generated continuously rather than in the halting turn-by-turn rhythm of older systems, a lag of only a few seconds, and synthetic output that holds onto the original speaker's intonation and pace. The engineering is real. It is also, increasingly, beside the point.

Lost in distribution

What Google built is no longer scarce. Apple shipped live translation through AirPods with iOS 26, in roughly nine languages, and switched it off entirely inside the European Union. Samsung, whose Galaxy AI now spans 22 languages, has been folding the same trick into its handsets since the Galaxy S24 and reckons it will be running on more than 400 million devices by the end of 2025. Meta has its Seamless models; a clutch of Chinese hardware firms have their own. Real-time speech translation, the stuff of science fiction within living memory, has in the space of two years become a checkbox every consumer-technology giant is now obliged to tick. The models differ at the margins, and none is a moat.