Making a product video 15 seconds long instead of 10 is only part of the change. For a working creative team, the more useful questions are whether an existing clip can be revised and how those revisions affect the bill. Google announced Gemini Omni 1.2 Flash on November 5, with general availability on the Gemini API and Google Cloud, longer ordinary generation and editing, and an effective 720p video-output price of about $0.07 per second. It is not an announcement of free 15-second video for every Google account.
The extra five seconds do not apply to every operation
Ordinary generation and editing now support 3–15 seconds, compared with a 10-second maximum for Omni 1.1. The launch announcement describes duration settings in one-second increments. Multi-turn editing lets users request changes to an existing clip, such as replacing a background, adjusting lighting or swapping a product, without describing the entire scene again.
Developers should not simply change every duration parameter to 15. Google Cloud's model documentation limits first-and-last-frame generation to 3–10 seconds. The new output added by an extension operation is also 3–10 seconds. These are different limits from ordinary generation and editing, so extension should not be treated as a way to add 15 seconds in one operation.
Google says the update improves consistency in characters, packaging text and camera movement, while acknowledging remaining problems with fast sports action, rapid script changes and complex physics. These are the vendor's descriptions, not results from our own testing or a reason to stop checking individual frames.
Seven cents per second is an output estimate, not the whole bill
The Gemini API pricing page offers no free tier for this model. Standard paid pricing is token-based: $1.50 per million text, image and video input tokens; $9 per million text output tokens; and $12 per million video output tokens, down from $17.50 for the previous version. Google's announcement describes the effective 720p per-second reduction as 30%.
At the documented 5,792 output tokens per second of 720p video, the video-output portion of a 15-second clip is approximately $1.04. That excludes input, text output and additional generations or revisions. It is not a promise that a finished advertisement costs $1.04. The pricing notes also explicitly describe upsampling for 1080p and 4K; those tiers should not be advertised as native 4K generation.
Our analysis: teams producing background and format variants of the same product should test iteration, not just longer duration. Compare total consumption per usable final clip rather than the per-second price of one successful attempt. Screening low-resolution drafts before increasing the resolution of selected clips is a manageable experiment, not a guaranteed saving.
API availability and app rollout are different things
Developers can inspect their account's access through the model entry in Google AI Studio. The announcement says 15-second video in the Gemini app is rolling out to Google AI Ultra subscribers. Flow supports scene direction and multi-turn editing. Google Vids is also listed, but its “no additional cost” wording must not be extended to free API use or unlimited generation: check the relevant account's product allowance.
There is also an availability discrepancy for business chat. The general announcement includes Gemini for business, while the specific access section of the Cloud blog says in-chat generation is coming soon. We therefore do not treat universal access in that chat interface as confirmed. Region, account eligibility and quotas still need checking against product documentation and the console; model GA alone does not establish access in a particular location.
Audio references have conditions too. The Cloud announcement says they require allowlisting, and the model card says the ability to change people's speech is currently restricted. Confirm both product permissions and rights to the material before preparing voice references or footage of real people.
Validate one clip before building a batch workflow
A creator can start with a product image they have permission to use, generate a short clip, and change just one variable, such as the background color. Check whether product shape, branding, motion or sound changed unintentionally. Developers should log the model ID gemini-omni-1.2-flash, operation, duration, resolution and actual usage, testing ordinary generation, first-and-last-frame generation and extension separately.
Existing 1.1 integrations also deserve regression testing. The Gemini API pricing page says requests to gemini-omni-1.1-flash will automatically route to 1.2 and use its rates from December 3, 2026. That is a Gemini API migration notice, not a promise covering every third-party platform. For the app entry point, see our Gemini tool page and check video access for your own account.