Google launched Guided Vision in Gemini Live on 1 October 2026, bringing real-time spoken descriptions of a phone camera’s view to compatible Android devices, according to its launch announcement. Built for people who are blind or have low vision, the feature lets someone share their camera during a conversation, ask questions about what it sees and hear directions for adjusting the shot when Gemini needs a better view.
Key points
- Guided Vision describes a shared camera view and takes follow-up questions in Gemini Live.
- Google says more than 1,000 Aira testers helped refine the feature.
- It is available on Android 9 and above where Gemini Live is supported.
- Google warns against using it for navigation or obstacle detection.
Gemini Live talks through the camera view
Guided Vision starts when the camera is shared in a Gemini Live session. Google says Gemini can describe the surroundings, answer a more specific question and give spoken directions if the camera is too high, too close or pointed off to one side. Those directions can ask for a slow pan to the right, a downward tilt or a step back. The camera’s frame sets the boundary of what Gemini can examine: a question about something outside the shot may first require moving the phone.
Reading the small print on a food package could become a conversation about the label rather than a single attempt to frame it perfectly. A description could help establish what is in view, a follow-up question could narrow the task to particular text, and a spoken cue could help bring the label into the shot if the camera missed it. Each step would depend on what the camera captures and how Gemini interprets it.
Google gives other examples that call for different kinds of visual information: reading an appliance dial, finding an earbud on the floor, locating pepper in a crowded cupboard and checking the colours of clothes. A dial calls for fine detail, while a lost object calls for a wider view of its surroundings. Guided Vision uses the same exchange of camera view, spoken description and follow-up questions for both kinds of task.
Apple’s VoiceOver Live Recognition offers comparable audio descriptions on iPhone and Vision Pro, The Verge reported. Google’s addition places that kind of assistance inside Gemini Live, where a question can follow the initial description and a spoken prompt can help bring the subject into view.
Aira testers helped refine Guided Vision
Google says it partnered with accessibility service Aira to train Gemini Live for conversations about real-world surroundings, drawing on tens of thousands of hours spent describing visual data. More than 1,000 members of Aira’s Trusted Tester network then helped stress-test and refine the model during daily routines, according to the company. Aira specialists also worked with Google’s teams on establishing and evaluating safety guardrails.
That work addresses a practical difficulty in asking questions of a camera: the answer depends not only on recognising an object or reading text, but also on knowing whether the image contains enough of the right detail. Google says Gemini gives spoken cues to pan slowly to the right, tilt down or step back when the camera needs reframing.
Google says testing and feedback also came from blind and low-vision communities across India, Brazil, Singapore, Indonesia and Japan, among other places. Guided Vision supports conversations in multiple languages. In Google’s announcement, Prashant Verma of the National Association of the Blind, India, described testing with blind and low-vision users across Indian languages as important to dependable assistance in daily life.
Sally Khoo, Deputy Director (Innovation) at SG Enable, connected the spoken camera guidance to a limitation of phones built around a visible screen. In a statement published by Google, she pointed to reading fine print and finding everyday essentials at home as tasks for which verbal descriptions could help.
Android 9 access comes with travel limits
Guided Vision is available on devices running Android 9 and above in regions and languages where Gemini Live is supported, Google says. In the Gemini app, it can be switched on through profile settings before the camera is shared in a Live session. It can also be launched through an Android Accessibility shortcut or selected from the Google TalkBack menu.
The shortcut offers several ways to start the feature without finding it in the Gemini app: Google lists a floating accessibility button, a two-finger swipe and a press of both volume keys. For TalkBack users, the route is a three-finger tap to open the TalkBack menu, followed by Guided Vision. These are access routes to the same camera-based conversation, rather than separate visual-assistance features.
Google warns that Guided Vision can make mistakes and is not intended for navigation, safe-travel guidance or obstacle detection. It is not a medical device, mobility aid or replacement for a white cane, the company says. Its examples concern examining what is nearby — such as text on a package or an object in a room — rather than deciding whether a route is safe to cross.
Google previewed Guided Vision in its September Android Drop as a forthcoming feature, Unite.AI reported. For the 1 October launch, Google says the Gemini app and Android system software should be updated to their latest versions before use.
