Gemini Omni 1.2 Flash: 15-second video clips at 7 cents a second, with no benchmarks yet

Google released Gemini Omni 1.2 Flash on November 5, 2026, an update to its video generation and editing model. The headline changes are clip length, up to 15 seconds from 10, and price, down about 30%. The post is Google's own; as of today we found no independent hands-on testing, so this piece separates what the documents prove from what they only assert.

If you are weighing Gemini Omni 1.2 Flash pricing against rivals, the useful starting point is the pricing table, not the launch copy.

What changed

According to Google's launch post, Omni 1.2 Flash generates and edits native video up to 15 seconds, with duration set in one-second steps from 3 to 15. It supports multi-turn editing, so a clip can be refined conversationally, for example swapping a product or changing the lighting, without regenerating from scratch. It accepts character, product or style images, short motion clips of up to three seconds and audio cues. Google says faces, garments and brand colors stay consistent across camera angles, and that speech has clearer lip-sync and multi-speaker conditioning. Camera moves, it says, come with "significantly fewer unwanted jump cuts."

Every sentence in that paragraph is a Google claim. The post is candid about limits: it says development continues on "fast sports action, rapid script changes and complex physics."

The price, checked against the pricing page

The post says 720p costs $0.07 per second, a 30% reduction from Omni 1.1. Google's pricing documentation backs that. Billing is by output tokens at 5,792 tokens per second of 720p video, which works out to about $0.07 per second, roughly $0.70 for a 10-second clip and about $1.04 for 15 seconds. The page lists other resolutions as:

• 360p: about $0.023 per second
• 1080p with upsampling: $0.104 per second
• 4K with upsampling: $0.209 per second

Omni 1.1 Flash is listed at about $0.10 per second for 720p, so the 30% cut checks out arithmetically. Google suggests drafting at 360p or 720p, fixing details with follow-up edits and upscaling only the keepers. That workflow matters because it changes the real cost: a 15-second finished clip at 4K is about $3.14 before any retries, and retries are the normal way people use generative video.

Where it ships

Google says the model is generally available in Google AI Studio and the Gemini API, in Google Cloud with Provisioned Throughput, in Flow, in Google Vids at no extra cost with a Google account or Workspace subscription, in the Gemini app for Google AI Ultra subscribers on a rolling basis, and in Google Ads. Outputs carry an imperceptible SynthID watermark. For teams on the older model, the pricing page says requests to gemini-omni-1.1-flash will be routed to 1.2 Flash on December 3, 2026 and billed at 1.2 Flash rates, so the price cut arrives automatically.

What is not established

RuntimeWire, an outlet that covered the release the same day, states that it found no independent testing and that the consistency claims are unverified. We found the same. The launch post offers examples, not scores, and it reports no human preference test, no comparison with rival video models and no failure-rate data. A cheaper model that drifts more on long camera moves can cost more per usable clip than a pricier one.

Two practical tests would settle most of it. First, run the same 15-second prompt with a reference character image ten times and count how many clips keep the face and garment intact. Second, price your real workflow, including edit turns, not a single generation.

How it fits Google's wider model push

Omni 1.2 Flash sits beside the frontier language releases we have covered, such as Gemini 4 Carbon and its claimed intelligence index score. Video is a separate product line with separate economics: the commercial pitch here is volume at low cost for advertising and short-form content, which is why Google lists Ads and Vids among the surfaces.

Working out the real cost

Using the pricing table, a finished 15-second clip costs about $1.04 at 720p, $1.56 at 1080p and $3.14 at 4K, before any retries. That is our arithmetic from Google's per-second rates, not a figure in the post. A draft at 360p costs about $0.35 for 15 seconds. The workflow Google suggests, drafting low and upscaling the keepers, therefore costs roughly $0.35 per attempt plus $1.56 for a 1080p final, which is far cheaper than iterating at full resolution.

If a team needs 20 usable 15-second clips and keeps one of every five drafts, that is 100 drafts at about $0.35, or $35, plus 20 finals at $1.56, or about $31, around $66 in total at list prices. The point is not the exact figure but that retries dominate the bill, which is why an unquantified consistency claim matters commercially. A model that keeps a character's face and garment intact on the first or second try is cheaper in practice than one priced lower per second that needs five tries, so the sticker price is only half of the comparison a buyer should make.

Our take

Take the price as fact and the quality as a hypothesis. A 30% cut and 15-second clips are real, documented changes that lower the cost of experimenting. They do not show that the output is better, and Google's own list of weak spots, fast action and complex physics, is where many generative video models struggle. If you buy video generation at scale, wait for a third-party comparison or run your own before committing budget.