Google's Guided Vision Turns Gemini Live Into a Real-Time Narrator for Blind and Low-Vision Android Users

Google has switched on a feature that turns the phone camera into a running conversation. Guided Vision, now rolling out inside Gemini Live on compatible Android devices, lets users share their camera feed and receive continuous spoken descriptions of whatever the lens sees - small print on packaging, an object buried on a crowded shelf, the layout of an unfamiliar room - and ask follow-up questions in the same conversation. It is aimed primarily at people who are blind or have low vision, and it works on Android 9 and above wherever Gemini Live is available.

What separates it from a one-shot photo scan is the dialogue. When the requested object is out of frame, Gemini talks the user into position - Google says the system offers "natural verbal prompts to help you reframe," asking the user to pan slowly to the right, tilt downward or step back - and then describes or reads out the details, like an expiration date on a specific item. Google's own launch demonstrations include finding a particular spice among crowded bottles and reading a menu in a dimly lit restaurant.

The build-out is notable for who was involved. Google partnered with Aira, an accessibility firm that provides remote visual interpreters for blind users, and says Aira specialists visually interpreted tens of thousands of hours of data for training, while more than 1,000 members of Aira's Trusted Tester network stress-tested the system. Testing explicitly pulled in communities in India, Brazil, Singapore, Indonesia and Japan rather than only English-speaking markets, and Aira also helped write the safety guidelines around the feature.

Getting to it does not require a new phone. Guided Vision can be enabled from the Gemini app's settings, set as an accessibility shortcut in the Android settings app, or launched with a three-finger tap from the TalkBack menu. Once on, the camera can be shared mid-conversation without switching apps, and the same thread keeps the context for follow-up questions. The positioning puts it alongside Apple's VoiceOver Live Recognition on the iPhone and Vision Pro, though Google's version leans on a live dialogue model rather than discrete recognition modes.

Google is unusually blunt about the limits. Guided Vision, the company says, is an assistive utility that can make mistakes: it is not a medical device, not a mobility aid, and not a replacement for a white cane, and it should not be relied on for navigation, safe-travel guidance or obstacle detection. No accuracy figures were published alongside the launch, and the feature needs a continuous connection and meaningful battery while the camera is shared - costs worth knowing before trusting it with something consequential like a medication label.

As a product matter, the launch is also a quiet piece of positioning. Assistive credibility is now part of how AI assistants differentiate: a vision feature built with a specialist partner, tested by the community it serves and shipped with honest limitations attached sets a bar that rival camera demos - which typically launched without either - will be measured against.