CONFIRMED
Google announced Gemini Omni 1.2 Flash on Thursday, November 5, extending video generation and editing to 15 seconds while cutting pricing by 30% compared with Omni 1.1. The advertised rate is $0.07 per second for 720p video. The change brings generative artificial intelligence closer to an everyday small-business need: producing short advertising variations without rebuilding every asset from scratch.
More room to demonstrate a product
Maximum duration rises from 10 to 15 seconds, with one-second increments available between 3 and 15 seconds. That creates room for a simple demonstration, a transition and a closing message within one generation. It does not automatically deliver a finished commercial. Scripting, selecting takes, reviewing details and completing the edit remain production work that businesses must budget for.
Savings extend beyond the headline rate
Users can refine the same clip through conversation, changing its background, lighting or aspect ratio, for example. Google also claims improved preservation of characters, clothing, products and visual references across frames. This matters for catalogs and campaigns, but supplier-selected examples do not replace independent testing of fidelity.
At the announced rate, generating 15 seconds at 720p costs $1.05 for video generation, before any other applicable charges, taxes or currency conversion. Ten attempts of that length cost $10.50 even if only one is approved. For a small Brazilian retailer, the useful metric is cost per usable asset, not simply price per second. A 30% reduction in the rate card does not guarantee an equivalent reduction in the complete campaign budget.
Where the update is available
The announcement states general availability through the Gemini application programming interface, or API, and Google Cloud. It also describes distribution through products including Flow and Google Vids. In the Gemini app, access is rolling out to Google AI Ultra subscribers. Multiple entry points do not mean immediate universal access: businesses should check plans, regions, quotas and the terms of each product before offering customers a service built on the model.
Outside the announcement, AI Tracker recorded the addition of gemini-omni-1.2-flash to the Gemini API model catalog on November 5. That provides external evidence of the model's presence, not proof of visual quality or operational savings. We did not run our own video-generation test for this report, so performance claims remain attributed to Google rather than independently validated findings.
The limit is what must remain accurate
Google itself acknowledges remaining challenges with fast sports action, rapid script changes and complex physics. E-commerce introduces another requirement: packaging, color, fabric and text must not change merely to make a scene more attractive. Every asset should be checked against the actual product. Outputs carry the imperceptible SynthID watermark, which does not remove editorial responsibility or the need for permission to use reference images.
Our assessment is that the meaningful advance combines enough duration for a short advertisement with incremental adjustments. Start with a small test set and compare approval rates and review time against the existing workflow. Expand only if the cost per valid asset falls. Cheap video that requires excessive correction is still expensive.
Sources
Google: Gemini Omni 1.2 Flash announcement; AI Tracker: model added to the Gemini API catalog.
