Google's Guided Vision brings real-time AI camera narration to Gemini Live on Android

Google has launched Guided Vision, a Gemini-powered feature that narrates whatever a phone's camera is pointed at in real time. According to The Verge, the feature went live in Gemini Live on compatible Android devices this week, and it is aimed primarily at people who are blind or have low vision, though Google frames it as useful for anyone who wants extra visual assistance in specific situations.

What the feature does

After sharing the camera inside Gemini Live, users can have Google's AI read fine print, describe their surroundings, locate or identify nearby objects, and get details about a specific item. The interaction is conversational rather than one-shot: The Verge cites the example of asking Gemini to read the expiration date on a product it has just helped find.

Google has also built in help with framing. When a user asks about something that is not currently in the shot, Gemini responds with audio cues meant to help line up the camera correctly.

The Verge draws a direct parallel to Apple's VoiceOver Live Recognition, which serves a similar audience on the iPhone and Vision Pro. Guided Vision is Google's counterpart on the Android side, arriving first through its Gemini assistant before spreading to the company's accessibility tooling.

Where to find it

The feature is not confined to the Gemini app. Per The Verge, Guided Vision is also available in Google TalkBack, and users can assign an accessibility shortcut to it in the Settings app on devices running Android 9 and above. The rollout follows the feature's announcement in a recent set of Pixel updates.

The limits Google spells out

Google is explicit about what Guided Vision should not be used for. The company cautions against relying on it for "navigation, safe-travel guidance, or obstacle detection," or as a replacement for a cane or mobility aid. That framing keeps the feature positioned as an assistive layer for reading, identification and description, rather than as a safety-critical system for moving through the world.

Why it matters

Real-time scene description is one of the most concrete accessibility applications of multimodal AI so far: it turns an ordinary phone camera into a conversational visual assistant. Because Guided Vision reaches back to Android 9, the pool of devices that can run it is large, and its presence in TalkBack places it inside a workflow that many blind and low-vision Android users already depend on. The launch also matches Apple's VoiceOver Live Recognition, putting comparable camera-narration capability on both major mobile platforms and raising the baseline of what ships by default. Google's own warnings, however, set expectations: this is a helper for tasks such as reading labels and finding objects, not a mobility aid, and its output should be treated as assistance rather than authoritative guidance.