Google launches Gemini Omni Flash 1.2: videos up to 15 seconds and a 30% lower price

The biggest pain point in generative video for products is not the quality of an individual frame, but control. You get an almost perfect clip, but the character changes jackets at the five-second mark, the camera moves in the wrong direction, and every edit requires generating the entire video again from scratch. Gemini Omni Flash 1.2 targets precisely these problems.

Google's Logan Kilpatrick announced the rollout of the update. Details beyond the announcement remain scarce, so below is a breakdown of what has been claimed and what users should verify for themselves.

What changed

The announcement lists five changes:

• video length increased to 15 seconds;
• more advanced cinematic controls for the camera viewpoint were added;
• editing through a multi-step dialogue is supported;
• visual drift has been reduced;
• the price is 30% lower than Omni Flash 1.1.

Each item may seem modest on its own. Together, they change the workflow: creating a clip becomes less of a lottery and more of an iterative process.

What this looks like in practice

Imagine you are making an ad for a delivery app. Previously, you would write a prompt, receive 5–8 seconds of footage, notice that the courier started to "drift" near the end, and regenerate the whole thing. Now you can build a 15-second scene and then say, "make the camera orbit slower" or "change the backpack to red," without losing everything else. Less drift means that characters and objects retain their identities for longer, which is critical for branded content.

Conversational editing is particularly important for products in which users generate the videos. Instead of a "prompt and hope" field, the interface can offer an editing chat similar to an image editor.

Pricing and economics

The 30% cost reduction is arguably the most practical change. Video generation remains expensive, and iterations multiply the cost. If your use case involves 3–5 edits per clip, a cheaper generation unit and the ability to edit instead of regenerating can reduce the bill significantly. One caveat: the announcement does not specify how much conversational edits cost or whether each edit is billed as a new generation.

Limitations

The claims of "less drift" and "advanced camera controls" currently come without figures or independent testing. A 15-second video is still a short clip: producing a complete story will require stitching scenes together and maintaining consistency between them. It also remains unclear how the model handles people, text within the frame, and complex physics.

What to do

If you already use Omni Flash 1.1, evaluate a migration using the same prompt set, compare drift across 10–20 typical scenes, and recalculate the budget with the discount included. If you are building a product with user-generated video, test multi-step editing as the foundation of the interface. If you are only beginning to explore video AI, the combination of 15-second clips and a lower price is a reasonable reason to run a pilot on a real task rather than on demo examples.

Bottom line: this is a practical release, not a revolutionary one. Google is improving predictability and pricing rather than the "wow factor," and those are precisely the factors that determine whether video generation makes it into production.