A complete 15-second vertical ad spot now costs about a dollar and five cents to generate. That is the number worth keeping from Google's Gemini Omni 1.2 Flash launch — more than the feature it is named for.
The model, generally available now on the Gemini API and Google Cloud, raises native clip length from 10 seconds to 15, addressable in one-second steps from 3 to 15. Price falls 30% from Omni 1.1 to $0.07 per second at 720p. Fifteen seconds at that rate is $1.05 before retries — the same order of magnitude as a stock photo, for a moving image with synchronised audio.
The second cost lever is workflow rather than list price. Google's pitch is to draft cheap at 360p or 720p, fix details with a follow-up conversational edit instead of regenerating the clip, then upscale the keeper to 1080p or 4K. Multi-turn editing covers background swaps, lighting, weather, product changes and aspect ratios across a full 15-second clip — exactly the loop where generative video has been burning money, since every small change used to mean a fresh generation.
Reference adherence is the substantive upgrade. Character, product and style images can be combined with motion clips of up to three seconds and audio cues, and the model is claimed to hold faces, garment drape, packaging typography and brand colours across camera angles. Google's own demo material — a 15-second virtual try-on sequence built from one model photo and three outfits, and a four-shot science explainer built from a single style frame with tracked callouts — is aimed squarely at commerce and advertising, and the quoted customers (Adobe Firefly, Figma Weave, Magnific, Monks, Havaianas via PORTO ROCHA, Whering, STAN) confirm it.
Speech and camera work improved too: more natural dialogue, tighter lip-sync, multi-speaker conditioning, and continuous FPV fly-throughs or orbit moves with fewer unwanted cuts — the Pantheon demo runs from the piazza to the oculus without one.
Two caveats Google states itself: the model is still being worked on for fast sports action, rapid script changes and complex physics, and every output carries a SynthID watermark. Distribution is broad — Flow, Vids, Ads, AI Studio, and 15-second clips in the Gemini app for AI Ultra subscribers.
The strategic read: at $0.07 per second with conversational editing, the constraint on AI video shifts from generation cost to review cost. The bottleneck becomes a human deciding which of forty variants is on brand — which is why the reference-adherence and consistency work matters more than the extra five seconds.
