Gemini Omni 1.2 Flash: 15-Second AI Video at $0.07 per Second

What happened

Google DeepMind has made Gemini Omni 1.2 Flash generally available, a video-generation model that now produces clips of up to 15 seconds. According to The Decoder, users can edit a clip step by step through conversation, and the model costs $0.07 per second at 720p, 30 percent less than its predecessor.

The pricing has a practical element too. Teams that draft at low resolution and upscale only the final version can reduce costs further, which makes the production workflow as important as the headline rate.

Taken together, the update combines longer output, conversational editing and a lower unit price. That combination moves AI video generation closer to the iterative, revise-and-refine process that many organisations already use for text and images.

Why it matters

For leaders in AI and digital transformation, the release marks a shift from experimental video output towards a repeatable production step. Longer clips mean fewer stitched segments, and conversational edits let people who are not specialist editors refine a result without starting again. The cost per second is the figure that will decide whether a pilot becomes a standing line item in a marketing, training or support content budget.

The experience implications follow from speed. Shorter gaps between a brief, a draft and a finished asset can change how quickly service and onboarding materials are tested and corrected. Organisations that treat video as a product to be iterated on, rather than a campaign to be commissioned, are likely to get more from tools like this.

By the numbers

• 15 seconds: maximum length of a generated video clip
• $0.07 per second: price at 720p resolution
• 30 percent: reduction in price compared with the predecessor model

The Renascence take

Most readers will focus on the 15-second ceiling, but the more consequential detail is the unit economics. Cheap, editable output changes who gets to make video and how often they revise it, and that affects service design as much as creative production.

The headline is about length, but the real shift is that revision has become cheap. Behavioural economics tells us that when the cost of a second attempt falls, teams test more options and stop defaulting to the first draft that looks acceptable. A customer-obsessed operator should use this window to run controlled tests on onboarding and help content, measuring whether shorter feedback loops produce clearer materials for customers rather than simply more of them.