Google is putting Gemini in a more practical place: the camera view. Its new Guided Vision feature uses Gemini Live to describe what a phone is seeing in real time, turning the camera into an interface for questions about the physical world.
Guided Vision is launching on compatible Android devices and can help users read small text, describe nearby surroundings, find or identify objects, and explain details on a specific item. For people with low vision, that could make an everyday phone camera more useful without requiring a separate visual-assistance app.
Gemini Live becomes a visual assistant
The important shift is not simply that Gemini can now see. Google is making the camera a conversational starting point. A user can point at an object, ask Gemini to describe it, and then continue with follow-up questions. If the feature identifies an item, the user could ask it to read an expiration date or provide more detail.
Guided Vision is also reported to work through Google TalkBack, while Android users can set up an accessibility shortcut on devices running Android 9 or later. Google’s system can provide audio cues to help users position the camera when the object they are asking about is not fully in view.
That makes the feature feel less like an occasional image search and more like an always-available layer between the user and the environment. The phone is no longer just capturing what is in front of it. It is interpreting the scene and offering a path to the next question.
The useful feature is also the risky one
Google is drawing a firm boundary around that interpretation. Guided Vision should not be used for navigation, safe-travel guidance, or obstacle detection, and it is not intended to replace a cane or other mobility aid.
That warning matters because the feature’s value depends on users trusting an AI description quickly. Reading a label or identifying an object is one kind of task. Judging whether a path is safe is another. The interface may feel conversational, but the underlying answer can still be incomplete, delayed, or wrong.
For Google, this is a meaningful step in normalizing AI as an ambient accessibility tool. For users, it offers a faster way to ask questions about the world around them. The next challenge is making sure the convenience of a voice answer never disguises the difference between assistance and dependable perception.