Google just made it cheaper and easier to generate convincing AI video—and the same corporation that decides what you're allowed to say online now controls the tools to fabricate what you see. Google's Gemini Omni 1.2 Flash, its AI video generation model, went generally available Thursday with clips up to 15 seconds long and a 30% price cut, according to TNW. It now costs roughly $0.07 per second of 720p video—about a dollar for a full 15-second clip, down from $0.10 per second in June.
That means anyone with API access can churn out photorealistic, editable video for pocket change. Users can now tweak a clip over multiple chat turns—swapping a product, changing a background, adjusting the lighting—without starting from scratch. The Verge reports the model also handles continuous camera movements and what Google calls "natural-sounding" dialogue. Google says faces, clothing, packaging text, and brand colors now stay more consistent across angles, with testers preferring the new version 62% of the time in internal comparisons.
Neither outlet spent much time on the obvious question: what happens when the company that already filters search results, demonetizes content, and throttles dissenting voices also holds the cheapest tool for manufacturing photorealistic video? Google says every clip carries an invisible SynthID watermark. But invisible watermarks are only useful if someone is checking for them—and ordinary Americans don't have the decoder. The watermark is a corporate comfort, not a public safeguard.
TNW noted that ByteDance's Seedance 2.5 can already generate 30-second clips, which means the AI video arms race is moving fast and the U.S. isn't alone in it. Google is rolling the model into Google Ads, Google Vids, the Gemini app, and Google Cloud for business customers. The ad integration is the tell: this isn't a creativity tool, it's a persuasion engine at scale.
Google admits the model still struggles with fast sports action, quick script changes, and complex physics. Those limits won't last long. The price will keep dropping. The clips will keep getting longer. And the people who decide what's "misinformation" and what's "harmful content" will be the same people selling the technology to fabricate both.
The real flashpoint isn't whether AI video is impressive—it is. It's who owns the means of fabrication, and whether the public will ever get a tool to tell what's real from what's made in a chat window for 70 cents.
