Gemini Omni 1.2 Flash makes longer video clips cheaper, but revision quality matters more

Google's Gemini Omni 1.2 Flash extends native video generation from ten to fifteen seconds and lowers the advertised 720p price to $0.07 per generated second. The more consequential promise is that creators can revise the same clip conversationally instead of discarding it whenever one detail needs changing.

The November 5 product announcement describes improved reference adherence, longer clips and multi-turn editing. A November 6 Google Cloud post confirms general availability on Cloud and adds access qualifications that matter when turning a demonstration into a workflow.

Fifteen seconds changes the unit of production

A clip can now run from three to fifteen seconds, specified in one-second increments. A longer native generation gives a short scene more room before it must be joined to another clip. That can simplify an explainer or product presentation, although it does not guarantee that all objects remain stable throughout the shot.

Google claims stronger consistency for faces, clothing, packaging text and colours across camera changes. It illustrates the claim with staged product and character examples. These are vendor demonstrations, not independent measurements of how often arbitrary references survive a difficult request.

The distinction matters for commercial use. A product video that looks polished but changes the shape of a fastening or a label can misrepresent the product. Better adherence is valuable only insofar as the details that matter remain accurate; a longer clip also provides more frames in which a mistake can appear.

The inexpensive draft is only part of the bill

At the advertised 720p rate, a fifteen-second generation costs $1.05 before considering other charges or subsequent work. That is simple multiplication of the published rate, not a quote for delivering a finished advertisement. Rejected attempts, edits and higher-resolution output can change the total.

Google describes drafting at 360p or 720p, then upscaling selected results to 1080p or 4K. Native duration and final output resolution are separate features. A 4K delivery option should not be described as proof that every draft is generated natively at that resolution.

Access differs between API and app surfaces

The two announcements do not describe every interface in identical terms. The main post presents a broad rollout across Google products, while the later Cloud article says direct creation in Gemini for business is coming soon. Developers should therefore distinguish confirmed API and Cloud availability from the appearance of a particular button in a chat application.

Google Vids is another supported route described by the company. Audio-reference access is specifically gated, so general availability of the model should not be taken to mean unrestricted use of every input type by every account.

Outputs carry Google's imperceptible SynthID watermark. That is a property of the generated video, not a substitute for checking whether a clip accurately represents a product or whether the user has permission to supply the reference material.

Revision is the feature to watch closely

Changing lighting or a background while retaining a useful performance could save more work than simply generating a few extra seconds. But the valuable test is whether the requested change stays local: a better background is little help if the object being advertised quietly changes too.

Google acknowledges remaining work on fast sports action, rapid script changes and complex physics. Omni 1.2 Flash offers a more flexible production unit, with a lower published starting price. Its practical value will depend on how much of a good first attempt remains intact after the corrections that real projects inevitably need.