Gemini Omni 1.2 Flash: $0.07/sec, 15-second product video

A 15-second product video from your own catalog photos now costs about a dollar to generate. Google rolled out Gemini Omni 1.2 Flash on November 5. It is generally available in the Gemini API and Gemini for business at $0.07 per second of 720p video, 30% below Omni 1.1. The price cut is useful. The bigger change for merchants is that Google now says the model keeps packaging text and brand colors consistent from the first frame to the last.

What actually happened

Per Google's announcement:

• Price. $0.07 per second for 720p, 30% lower than Gemini Omni 1.1. A full 15-second clip is about $1.05 before upscaling.
• Longer clips. Native generation and editing up to 15 seconds, up from 10 in Omni 1.1. You set the duration in one-second steps from 3 to 15 seconds.
• Reference adherence. You can combine character, product or style images, motion clips up to three seconds, and audio cues. Google says faces, garments, packaging typography and brand colors stay consistent across camera angles. It names virtual try-on, catalog spins and branded ads as uses.
• Multi-turn editing. You can change one clip turn by turn, such as the background, the product, the lighting or the aspect ratio, instead of generating it again.
• Draft, then upscale. Draft at 360p or 720p, fix details with edits, and upscale the keepers to 1080p or 4K.
• Production status. This is the first generally available Omni release. It runs on Google Cloud with Provisioned Throughput. Every output carries an invisible SynthID watermark.

Google quotes early users including Adobe Firefly, Monks, Figma Weave, and Whering, a virtual try-on app.

Why it matters for your business

When we covered Omni 1.1, we said most of the budget dies in throwaway takes. Multi-turn editing goes after that directly. If one edit fixes the label, you do not pay to generate the whole clip again. Make one approved clip per SKU, then edit it into a 9:16 social cut, a holiday background and a sale variant.

Do the math on your catalog. A store with 200 SKUs and one 15-second clip each is about $210 in generation cost. Add three iterations per SKU and the bill is still under $1,000. At that price, compute is not what limits you. Review is.

Test where generative video fails merchants: text on your product. "Typography stays consistent" is Google's claim from curated demos. Run your 20 hardest SKUs, the ones with small print, nutrition panels or exact color specs, through the model before you automate anything. A clip that changes your label is worse than no clip, and on regulated products it is a compliance problem. Keep a human approval step, and log the prompt and reference set for each approved clip so you can render it again.

Sources: Google Blog, Google for Developers on X.