Google's Gemini Live has just taken a major leap forward — it can now see. Rolling out to Pixel 9 series (including the new Pixel 9A) and Samsung Galaxy S25 devices, this new feature transforms the AI assistant into a real-time visual companion, capable of identifying objects, providing context, and even helping you navigate your environment.

In a hands-on test, Gemini Live proved to be more than just an upgrade to Google Lens. It offers a conversational, interactive experience — spotting household items, deciphering artwork origins, and even helping users locate misplaced scissors next to a packet of pistachios. The interaction feels natural, moving beyond rigid commands to something that resembles casual human conversation.

Google showcased Gemini Live in its April 2025 Pixel Drop and launched a dedicated page on the Google Store, while users can simply activate their camera and start chatting with the AI to explore its capabilities.

Pushing Boundaries: Beyond Simple Object Recognition

Gemini Live builds upon Google’s Project Astra, a significant leap in generative AI unveiled last year. It aims to blend the digital and physical worlds seamlessly. Unlike previous assistants, Gemini Live provides real-time feedback, contextual insights, and can even translate text it spots — perfect for bustling environments like train stations or art galleries.

In real-world tests, Gemini Live impressed early adopters. From accurately identifying chamomile and dianthus flowers to recognizing limited-edition gaming collectibles, the AI assistant showcased moments of brilliance. At an art gallery, it correctly identified a tortoise on a cross and translated accompanying kanji, stunning onlookers with its instant expertise.

However, it wasn’t flawless. Extended sessions revealed that Gemini occasionally struggled with more obscure items, especially horror-themed collectibles. Interestingly, some AI "sessions" performed better than others. Restarting conversations or revisiting older chats improved accuracy, hinting at evolving context management in the system.

Hits, Misses, and Hints

The AI shined when it came to popular and recent items, like pinpointing a collectible from Destiny 2, complete with event details. Yet, it sometimes tripped over niche objects, confusing them with others based on previous context, like frequently misidentifying items from the Silent Hill franchise.

One workaround: providing hints. Guiding Gemini with contextual clues often nudged it toward the right answer — though not always. And when it got stuck, restarting the session sometimes helped, but not consistently.

Despite occasional glitches, Gemini Live demonstrates huge potential. Its ability to meld conversational AI with visual recognition could redefine everyday interactions with our devices. Whether you're lost in a busy station, exploring a museum, or just looking for your keys, Gemini Live promises a future where your smartphone isn’t just smart — it’s observant, responsive, and interactive.

Google has yet to share deeper technical details about how Gemini Live functions behind the scenes, but early experiences suggest an exciting road ahead for visual AI assistants.