Google launches Gemini Omni 1.2 Flash: 15-second video, 30% cheaper
Google released Gemini Omni 1.2 Flash on November 5, 2026, a video generation model that makes clips up to 15 seconds long, up from 10 seconds in Omni 1.1. It is generally available on the Gemini API and in Google Cloud for the first time, and Google says it cut the price by 30%, to $0.07 per second for 720p video. For developers, that makes longer, cheaper AI video a production API call rather than a preview experiment.
What’s new
Everything below comes from Google’s announcement, written by Anish Nangia, a product manager at Google DeepMind.
- Longer clips. You can generate and edit videos natively up to 15 seconds, and set the duration in one-second steps from 3 to 15 seconds.
- Lower price. Google dropped the price 30% from Gemini Omni 1.1, to $0.07 per second for 720p video. The cost-saving workflow it suggests: draft at 360p or 720p, fix details with a follow-up edit instead of regenerating, then upscale the best result to 1080p or 4K.
- Better reference adherence. You can guide a generation with character, product or style images, motion clips of up to three seconds, and audio cues. Google says faces, garments, packaging typography and brand colors stay consistent across camera angles.
- Multi-turn editing. You refine the same clip turn by turn, for example swapping the product or background, changing lighting or weather, or switching the aspect ratio, and this works across full 15-second clips.
- More consistency. Google claims less visual drift or morphing between the first and last frame, including when you interpolate between a first and last keyframe or extend a clip.
- Speech and camera control. The model produces more natural dialogue, clearer lip-sync and distinct multi-speaker conditioning. Google also says complex continuous camera moves, such as FPV flythroughs and orbits, now come with significantly fewer unwanted jump cuts.
- Provenance. Outputs carry an imperceptible SynthID watermark, and Provisioned Throughput is available.
Google lists limits too: development is ongoing on fast sports action, rapid script changes and complex physics.
Availability
Omni 1.2 Flash is generally available on the Gemini API and Gemini for business, backed by Google Cloud. Google says this is the first time Gemini Omni is generally available.
- Developers: prototype in Google AI Studio, then use the Gemini API developer documentation and pricing page.
- Businesses: try it in the Google Cloud console and build with the API.
- Everyone else: direct scenes and multi-turn edits in Flow, generate 15-second clips in Google Vids at no additional cost with a Google account or Google Workspace subscription, and use Google Ads for video campaigns. In the Gemini app, 15-second video is rolling out to Google AI Ultra subscribers.
What it means for developers
The practical change is the workflow. Because edits happen in conversation, an app can keep one generated clip as the base and produce ad variants (a new background, a different product, another aspect ratio) without re-briefing the model each time. Google explicitly pitches this at ad variants for different audiences and placements.
The 3 to 15 second duration control in one-second steps also matters for budgeting. At the quoted $0.07 per second for 720p, a full 15-second clip works out to about $1.05 before any upscaling, and a 5-second draft to about $0.35. That arithmetic is ours, based on the announced rate; check the pricing page for 360p, 1080p and 4K rates, which the announcement does not list.
If you build on video models, plan for the reference inputs: images for characters, products and styles, short motion clips and audio cues are all supported inputs, so product catalogs can feed directly into generation. Google’s own examples include virtual try-on clips, catalog spins and branded video ads.
How it compares
Against its predecessor, the changes are measurable: 15 seconds instead of 10, and a price 30% lower. Google also says Omni 1.2 Flash matches reference images more closely than Omni 1.1 in its side-by-side examples. These comparisons are Google’s own; the announcement does not include independent benchmarks, so test it on your own prompts before committing a pipeline.
If you are wiring a model like this into an app, our earlier post on building AI-powered apps with Vercel’s framework covers the app-side plumbing.