On November 5, 2026, Google began the rollout of Gemini Omni 1.2 Flash — an updated version of its model for creating and editing video. The company's official blog and Cloud Blog confirm that the model is now available through the Gemini API, Gemini for Business, and a number of consumer products.
The key changes are an increase in clip length to 15 seconds versus 10 seconds in version 1.1, a price reduction of 30% to 0,07 dollars per second of video at 720p resolution, and the introduction of multi-step conversational editing. The user can refine details one at a time without recreating the entire clip.
The conversational editing mechanism allows changing the background, lighting, objects, or camera angle within a single clip. This is fundamentally different from previous approaches, where each change required a new generation from scratch. The model better preserves the consistency of characters, objects, and brand elements when using references — images, short videos, and audio.
Google's methodology emphasizes workflow: a draft at low resolution, edits in dialogue, then upscaling to 1080p or 4K. This scheme reduces costs and speeds up iterations, which is important for mass production of advertising clips, virtual try-ons, and explainer videos. However, the official materials lack detailed benchmarks against competitors or independent third-party evaluations.
In the video-generation landscape, this is a step away from single-step models like early Sora or Luma Dream Machine. Google is betting on integration with the Gemini ecosystem and conversational control, which brings generation closer to the familiar editing process. Compared with Omni 1.1, there is noticeable progress in adherence to references and in the naturalness of speech with lip-sync.
For developers and businesses, the price reduction and the ability to precisely control length from 3 to 15 seconds open up new scenarios. Previously, the limit of 10 seconds constrained complex narratives; now it is possible to create full-fledged advertising spots or product explainers in a single pass with subsequent edits.
The question of long-term consistency over longer sequences and generalization to complex scenes with many moving objects remains open. Independent tests have not yet been published, and real-world production performance will depend on specific cases.
The update shows that Google is focusing not only on generation quality but also on practical integration into workflows, bringing video AI closer to an everyday-use tool.
