Google has made Gemini Omni 1.2 Flash, its video generation model, generally available, with clips of up to 15 seconds and a price of $0.07 per second at 720p. Anish Nangia, a product manager at Google DeepMind, announced it on the Google blog on 5 November 2026. The price is 30% below Gemini Omni 1.1.
This is the follow-up to the story AI Insiders ran when Omni Flash first reached developers through the Gemini API as a public preview. That version stopped at 10 seconds, cost $0.10 per second, and came with a list of admitted gaps: no audio references, no scene extension, and shaky character consistency across scene changes. Version 1.2 goes after most of that list.
Clips can now run from 3 to 15 seconds in one-second steps, which Google says is enough for a product explainer or a full social ad spot in a single generation. Google’s cost advice is to start with low-resolution drafts at 360p or 720p, correct small errors through an edit request rather than a fresh render, and upscale only the keepers to 1080p or 4K. Google quotes a price only for 720p, so what a finished 4K clip costs is left for customers to work out from the pricing page.
The model also takes more direction from reference material than before. Developers can combine images of a character, a product, or a style with audio cues and motion clips that run no longer than three seconds. Google says the model keeps faces, clothing, lettering on packaging, and brand colors steady as the camera moves. Its showcase example pairs a single photo of a model with front and back views of three outfits and returns a 15-second try-on sequence. Any retailer with a catalog of product photography will see the pitch.
Editing now runs as a conversation across the full 15 seconds. A user can swap a product or background, change the lighting or weather, or switch the aspect ratio, then keep going turn after turn. In Google’s demo, a clip of two armored characters moves through five backgrounds, starting at a striped pavilion tent and ending at a stone castle, with the performances and framing intact. For agencies, that turns one expensive generation into a set of ad versions for different audiences and slots.
Speech and camera work also changed. Google describes more natural dialogue, clearer lip-sync, and separate voices for multiple speakers, plus long continuous moves such as an FPV drone flight through the Pantheon’s dome with no jump cuts. All of these examples are demos Google picked, and the post gives no measured rate of visual drift. Google itself lists sports action at speed, quick script rewrites, and complicated physics as work still in progress.
For business buyers the label matters most. Google says Gemini Omni is generally available for the first time, running on Google Cloud infrastructure with Provisioned Throughput (reserved capacity for steady workloads), and outputs carry a SynthID watermark that people cannot see. Access is spread across Google AI Studio, the Google Cloud console, Flow, Google Vids, and Google Ads. Vids gives Google account and Workspace holders 15-second clips at no extra charge, and the Gemini app is adding the feature for Google AI Ultra subscribers.
At $0.07 a second, a 15-second draft at 720p comes to $1.05, so a rejected take costs about a dollar. That is cheap enough to change how an ad team budgets for iteration, and the 15-second ceiling gives it room to cover a whole social ad rather than a fragment of one.
Reported by the Google blog on 5 November 2026.
