Guided Vision gives Gemini Live a camera for the real world

Google has released Guided Vision in Gemini Live on Android, turning the assistant into a spoken pair of eyes for people who are blind or have low vision. The tool was built with the accessibility community, previewed earlier this month, and from today is rolling out on compatible Android devices.

How it works

Share your camera inside a Gemini Live session and the assistant describes what it can see, answers follow-up questions about it, and speaks up when the picture is not usable. If the phone is pointed too high, too close or off to one side, Guided Vision asks you to pan slowly to the right, tilt down or step back so it has enough context to answer.

Google is aiming at everyday tasks where a quick visual check decides the outcome: reading a nutrition label, working out which cycle a washing machine dial is on, scanning a menu in a dim restaurant, finding a dropped earbud, checking whether a striped shirt goes with a particular pair of trousers, or getting a description of a room you have just walked into.

Trained with Aira's human interpreters

To teach Gemini Live conversational, real-world context, Google partnered with Aira, the visual interpreting service, for tens of thousands of hours of interpretation data. More than 1,000 members of Aira's Trusted Tester network put the model through daily routines, while Aira specialists worked with Google's teams as subject-matter experts to set and assess the safety guardrails.

Descriptions, reframing cues and answers were also tested with blind and low-vision communities in India, Brazil, Singapore, Indonesia and Japan, and Guided Vision holds a conversation in the user's preferred language rather than a fixed one.

Three ways to switch it on

Open your profile settings in the Gemini mobile app and turn on Use Guided Vision in Live, then start Gemini Live and tap to share the camera. Android users can also set an Accessibility shortcut under Settings, Accessibility, Vision assistance, Guided Vision, using the floating accessibility button, a two-finger swipe or both volume keys, and TalkBack users can reach it from the TalkBack menu with a three-finger tap.

What Google says it is not

Google describes Guided Vision as an assistive utility that can make mistakes, not a medical device, mobility aid or a replacement for a white cane, and not for navigation, safe-travel guidance or obstacle detection. Users are told to keep relying on established mobility aids. It runs on Android 9 and above, in the regions and languages where Gemini Live is supported, on an up-to-date app and system software.

Our opinion

The most consequential detail in this launch is where the training data came from. Google bought tens of thousands of hours of human visual interpretation from Aira and put its testers and specialists inside the process, which is a very different proposition from releasing a camera feature and inviting accessibility advocates to react to it. The second decision worth noting is the framing: Guided Vision is sold as help with labels, dials and dropped earbuds, and Google draws the line at navigation and obstacle detection before anyone else has to draw it. That is the correct line, because a model that is confidently wrong about a moving car is not a bug report, it is a safety incident. Whether the tool earns its place in daily routines will show up in the reframing cues, which are the part users cannot easily verify for themselves.