Gemini Omni 1.2 Flash Cuts AI Video Pricing 30%

I do not make videos for a living. I close books, chase receivables, and build Excel models. So when Google shipped Gemini Omni 1.2 Flash on 5 November 2026, my first question was not “how good is the video?” It was “does this replace the freelancer I pay for product clips?”

Here is the short version, in plain numbers: a 15-second 720p clip now costs about $0.07 per second on the Gemini API. That is roughly $1.05 for a finished 15-second clip. Google says the price is 30% lower than Gemini Omni 1.1. For a finance team, that is the whole story in one line.

What actually changed

Three things, and I will keep them practical:

• Longer clips. Native video generation went from 10 seconds to 15 seconds, and you can set the duration in one-second steps from 3 to 15. A single generation can now hold a full product explainer instead of half of one.
• Cheaper drafting. You can draft at 360p or 720p, fix a detail with a follow-up edit, then upscale only the final version to 1080p or 4K. You stop paying for full-quality re-renders every time something is wrong.
• Better consistency. You can feed character, product, or style images plus short motion clips and audio cues, and the model holds details — faces, packaging text, brand colours — across camera angles. Google’s example is a 15-second virtual try-on built from one model photo and front/back shots of three outfits.

There is also multi-turn conversational editing. Instead of regenerating from scratch when a logo drifts, you tell it what to change and it keeps the rest. Anyone who has rebuilt a spreadsheet from a broken template will recognise how much time that saves.

Why a chief accountant should care

Our marketing spend is small and lumpy. A product clip for a trade show, a short clip for a Facebook ad, a “how to read your statement” explainer for customers. Each of those used to be a purchase order to an agency, a two-week wait, and a revision thread.

At $0.07 per second, a 15-second draft costs about a dollar. I can generate four versions in the time it takes to write the brief. If none of them work, I have lost four dollars, not four hundred. That changes who gets to start the project — suddenly it is me, not a vendor.

The accounting angle is straightforward too. Video is now a small, predictable variable cost rather than a fixed project cost. When I budget marketing for next quarter, I can put a number per clip in the sheet and defend it. Before this, I could only put a range and hope.

Where it still costs money

I am not going to pretend this is free. Three things to watch:

• Rejected generations still bill. Cheap is not the same as free. A 3-second test is pennies, but a week of experimenting is a real line item. I now put AI video in its own expense category so I can see it.
• Upscaling is extra. The $0.07 per second is the 720p number. If you deliver in 4K, the final render costs more. Budget the delivered version, not the draft.
• Nobody reviews content for free. A wrong claim about pricing or a product benefit is still my name on it. The model does not know our warranty terms. I check every frame the way I check every invoice.

What I am doing this week

I have one concrete task queued: a 15-second clip of our range for the next customer email. I will generate three drafts at 360p, pick one, fix the label on the product, then upscale. Total expected cost: under five dollars. If it is not usable, I will say so in the next post instead of pretending it worked.

The bigger lesson I keep relearning this year: the useful question is never “is the model impressive?” It is “what does one finished unit of work now cost, and who has to approve it?” Gemini Omni 1.2 Flash moves both numbers in a direction I like.

Source: Google blog, “Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips”, published 5 November 2026. Pricing and availability as stated in that announcement — check the current Gemini API price list before you budget.