Google’s Gemini Omni 1.2 Flash generates 15-second AI video clips at $0.07 per second, according to the listing that surfaced on Product Hunt. That works out to roughly $1.05 for a single 15-second generation. The headline is not the model. It is the number attached to the second, and what that number does to everyone else selling generated motion.

The brief on Omni 1.2 Flash is thin: a 15-second ceiling and a per-second price. No resolution is published in the source, no frame rate, no audio spec, no mention of whether the output carries a SynthID watermark, which Google has attached to its generative media since 2023. Those gaps matter, and we will come back to them. What the listing does establish is a unit of account. Video generation now has a public price per second, and it is seven cents.

The per-second price is the product

For most of the past two years, AI video has been sold in the wrong units. Labs quote credits, subscriptions, or “generations per month,” which tells a buyer almost nothing about marginal cost. A creator who wants to know what it costs to render a 90-second spot has had to reverse-engineer the answer from a credit ledger. Per-second pricing ends that. It is the pricing model a production house actually budgets against, the way a studio budgets against footage rates or a cloud team budgets against GPU-hours.

At $0.07 a second, a minute of generated video costs about $4.20. Ten minutes costs $42. A hundred 15-second social cuts cost roughly $105. Those are numbers a small agency can put in a spreadsheet without a finance team. That is the shift. The model is being sold like a utility, not a subscription.

The comparison that matters is not against other video generators. It is against the cost of shooting. A single day of a modest commercial production, with crew, talent, location, and edit, runs into the thousands before anyone touches a camera. Omni 1.2 Flash does not replace that production. It replaces the animatic, the mood board, the pitch reel, the thing you make to win the budget for the real shoot. That work was always the cheapest part of the pipeline and the most tedious. Seven cents a second makes it nearly free.

What Google is actually doing here

Google is not first to cheap video. It is first to make the price legible. The “Flash” suffix is the tell: this is a tiered product, and Google is positioning the cheap, fast tier as the default entry point rather than the premium one. The company has run this play before with its language models, where a Flash tier sits below a Pro tier and absorbs the high-volume, low-margin traffic. Video is following the same curve, roughly a year behind text.

There is a strategic logic to selling the second, not the clip. A per-second price scales linearly with output, which means Google captures more revenue as a customer’s usage grows without renegotiating anything. It also makes the product easy to undercut and easy to compare, which is a strange thing to do voluntarily. Unless you are confident you can win on cost. Google owns its own silicon, its own data centers, and its own model stack, and it has been building custom accelerators for AI workloads for years. A vertically integrated lab can afford to set a price its competitors cannot match and still make margin.

That is the bet. Whether the margin holds is a separate question, and one the listing does not answer. Inference on video is expensive: a 15-second clip at any reasonable resolution is a large amount of compute, and seven cents a second is aggressive for the output. Either Google has found real efficiency gains in its video stack, or it is buying market share. Both are plausible. Only one is sustainable.

The gaps the listing leaves open

A price without a spec is a marketing claim, not a product sheet. The source does not say what resolution Omni 1.2 Flash delivers at $0.07 a second. It does not say whether that price covers 720p, 1080p, or something lower. It does not say whether audio is included, whether the model accepts image or video conditioning, or whether the 15-second limit is a hard cap or a soft one that can be chained into longer sequences. Each of those variables moves the real cost per usable second by a large multiple. A clip that needs three attempts to get right costs three times the sticker.

The watermark question is the one to watch. Google has put SynthID into its image and video generation tools, and the EU AI Act’s transparency obligations for synthetic media are phasing in through 2026. A model sold at commodity prices, aimed at high-volume social output, is exactly the kind of tool regulators will want provenance metadata on. If Omni 1.2 Flash ships without it, that is a policy story waiting to happen. If it ships with it, Google gets to set the default for how cheap generated video gets labeled, which is a quieter kind of power.

The model is being sold like a utility, not a subscription. That is the shift, and it resets the floor for everyone else.

What this does to the video-gen field

The competitive set for AI video has been defined by quality demos: longer clips, better physics, more coherent motion. Omni 1.2 Flash changes the axis. If a 15-second clip costs about a dollar, the question stops being “which model looks best” and becomes “which model looks good enough at this price.” That is a quality-per-dollar contest, and it favors whoever has the cheapest compute, not whoever has the best demo reel.

For startups in the space, that is uncomfortable. A company that raised on the promise of premium video quality now has to justify a price multiple against a Google tier that undercuts it by an order of magnitude on the low end. The defensible position moves upmarket, toward longer clips, higher resolution, better control, and the production-grade features that a 15-second commodity model cannot touch. The bottom of the market just got commoditized, and the bottom is where a lot of the volume lives.

For builders, the practical consequence is simpler. Video generation is now a line item you can price into a product without a partnership deal. A tool that generates a personalized 15-second clip on demand, at roughly a dollar a render, is economically viable in a way it was not when the same output required a negotiated enterprise contract. The interesting products will not be video generators. They will be the things that use cheap video the way apps used cheap text.

What to watch

The next data point is the spec sheet. Resolution, audio, conditioning, watermarking, and whether the 15-second cap can be extended. Until those are published, $0.07 a second is a floor, not a quote, and the real cost per usable clip sits somewhere above it. Watch whether Google publishes the details or lets the price do the talking. The price is loud enough that it may not need to say much else.

{/* TODO: confirm resolution, frame rate, audio support, and whether SynthID watermarking is applied to Omni 1.2 Flash output; source listing does not specify. /} {/ TODO: verify the $0.07/second figure is the current public list price and not a promotional rate; source is a Product Hunt listing, not a Google pricing page. */}