Point, hold, ask, get an answer.
For years, the gap between seeing and knowing has required a pause — a moment to type what the eyes already understood. Google Lens has now closed that gap, allowing users to speak their questions directly into the act of looking, powered by Gemini AI's contextual intelligence. The update, rolling out this week, transforms visual search from a two-step tool into something closer to a continuous, natural dialogue between human curiosity and machine understanding.
- The friction of switching from camera to keyboard — however small — has long interrupted the flow of visual discovery, and Google is now eliminating that pause entirely.
- By holding the shutter button and speaking aloud, users can ask layered questions about whatever their camera sees, from identifying blueberry counts to predicting a pumpkin's harvest timeline.
- Gemini AI powers the responses, meaning answers are not mere labels but contextual, conversational replies tailored to the specific question asked about the specific image.
- The rollout is moving quickly across Android devices, with users on older Lens versions needing only an update to unlock the capability.
- This change arrives just one week after Google added image provenance tools to Lens, signaling a deliberate push to make visual search feel less like a utility and more like an intelligent companion.
Google Lens, the image recognition tool that has helped people identify plants, decode foreign signage, and track down products spotted in the world, has gained a voice. Users can now hold down the camera shutter button and ask questions aloud, replacing the need to type follow-up queries after capturing an image.
The feature builds on multisearch, a capability Google introduced two years ago that let users photograph something and then ask questions about it — but only by typing. The new voice layer removes that interruption. Android journalist Mishaal Rahman demonstrated it by pointing his phone at blueberries and asking how many were in the frame; the answer surfaced naturally in a search result. A backyard pumpkin plant yielded a similarly fluid exchange: one voice query returned the accurate answer that fruiting takes 90 to 120 days.
The responses are generated by Gemini, Google's AI system, making them contextual rather than purely identificatory — users receive information shaped by what they asked, not just a name for what they saw. The rollout is moving swiftly; updating to the latest version of Google Lens should make the feature available on most devices. The update follows Google's recent addition of image provenance tools to Lens, part of a clear effort to evolve visual search from a discrete tool into something that feels like an ongoing conversation.
Google Lens, the image recognition tool that has spent years helping people identify plants, read signs in foreign languages, and hunt down products they've spotted in the wild, just got a voice. Starting this week, you can hold down the camera button and ask questions out loud instead of typing them in.
The feature builds on something Google introduced two years ago called multisearch—the ability to take a picture of something and then ask follow-up questions about it. Before now, those questions had to be typed. You'd photograph a dress, Google would find it for sale, and then you'd type "green" to narrow down the color. It worked, but it required a second step, a pause, a shift from looking to typing.
Now you just hold the shutter button and speak. Android journalist Mishaal Rahman demonstrated it by pointing his phone at a plate of blueberries and asking, "How many blueberries are in the picture?" The answer appeared in a Google search result. The interaction felt natural—point, hold, ask, get an answer.
Google has been developing this capability since earlier in the year, and the rollout appears to be moving quickly. The feature showed up on a Google Pixel 8 when tested, and it performed as advertised. A pumpkin plant growing in a backyard became the subject of a voice query: "How long does it take this plant to grow fruit?" The system identified the plant and returned the correct answer—90 to 120 days until fruiting. Without the voice feature, that same question would have required typing a separate search after Lens identified the plant. Now it's one continuous motion.
The voice queries are powered by Gemini, Google's AI system, which means the answers aren't just image identification—they're contextual and conversational. You're not just getting a label for what you're looking at; you're getting information about it, tailored to what you asked.
If you don't see the feature yet on your device, the reason is likely that you're running an older version of Google Lens. Updating to the latest version should make it available. This rollout comes a week after Google added its "About this image" feature—a tool that provides context about where an image came from and whether it's been altered—to both Circle to Search and Google Lens itself. The company is clearly moving fast to make visual search feel less like a tool and more like a conversation.
Citações Notáveis
Speak now to ask about this image— Voice prompt in Google Lens
A Conversa do Hearth Outra perspectiva sobre a história
Why does holding a button instead of typing matter so much? Isn't it just a different input method?
It's about friction. Right now, if you photograph something and want to know more, your brain has to switch from visual mode to typing mode. You have to think about what words to use. With voice, you just ask the question the way you'd ask a person standing next to you.
But voice search has been around for years. What's different here?
The difference is context. Before, voice search was separate from image search. You'd take a picture, get a result, then start a new voice search. Now the voice question is attached to the image. The AI knows what you're looking at while you're asking about it.
So it's more like a conversation with the image itself?
Exactly. You're not searching the web for information about blueberries in general. You're asking about the specific blueberries in front of your camera. The system has already seen them, identified them, and now it's answering a question about that particular thing.
How accurate is it? Can it actually count blueberries?
In the demonstration, it counted them correctly. But the real power isn't in counting—it's in questions like "how long until this grows fruit" or "what color is this dress available in." Those are the queries that would have been annoying to type out.
What happens next? Is this the end of typing into search?
Not at all. But it's another step toward search feeling less like a tool you use and more like a sense you have—point your camera and ask, and get an answer. That's the direction Google is clearly moving.