AI&Design
All change articles

Gemini image · Announcement · Official source

Gemini Live turns a phone camera view into an image edit

Jun Ishii pointed Gemini Live at a pair of earbuds and asked in Japanese for a blue version. About 30 seconds later, the app returned an image, according to his June 6 Jetstream walkthrough. The report pairs the spoken request with two interface screenshots. An object already on a desk became the reference for a color study without first taking and attaching a still photograph. The result is an in-app preview. The published record leaves color accuracy, design approval and delivery open. Jetstream’s June 6 walkthrough

Google described this camera-to-image interaction in its May 19 Gemini Apps release notes. The entry says a user can show Gemini what they see to create new images with Nano Banana while talking. It is part of a broader update that integrates Live with chat, shows visual responses and connects to other apps. Google says the wider update was rolling out globally on Android and iOS. The entry gives no activation date for individual accounts. May 19 Gemini Apps release notes

Early June coverage gives the interaction a second date. Jetstream and Impress Watch date Google’s notice about direct image creation and editing in Live to June 5 in the United States, June 6 in Japan. Impress Watch documents several paths. A user can ask for an image from scratch or revise an existing image by voice. On a phone, the shared camera can supply the scene to change. Those are paths inside the Gemini app. A developer building an integration with the Gemini Live API needs separate evidence for input handling and how an image tool is invoked. June availability and mobile paths · Gemini Live product overview

What the earbud example establishes

The operator supplied a visible object and one modification, a color. The article documents a single approximate wait with no stated clock boundaries, repetitions or equivalent run that attaches a photograph in ordinary chat. It also omits the device, app build and image model revision. A video feed brings viewpoint, lighting and background into the task alongside the spoken instruction. Comparing the result with a packaging or color specification would require the source and exported output, which are unavailable. The evidence covers a quick concept preview for a specific task. Jetstream’s inputs and screenshots

The signed-out Reddit search for May 12 through June 2 surfaced no dated original with an attributable camera-to-image result. Later public recaps describe the interaction, and their accessible submissions displayed relative dates. The June 6 operator walkthrough falls after the primary discussion window and supplies dated follow-up for the app rollout. Later organization recap · Later Reddit coverage