Gemini Omni 1.2 Flash goes GA and cuts the cost of generative video
On November 5, 2026, Google made Gemini Omni 1.2 Flash generally available on the Gemini API and Gemini for Business, raising native video clips to 15 seconds, adding multi-turn conversational editing, and cutting the price 30% to $0.07 per second at 720p. For teams producing AI video, the move to 15 seconds and the price drop change the cost-value equation of going to production.
November 5, 2026. Google announces the general availability of Gemini Omni 1.2 Flash on the Gemini API and Gemini for Business. 15 seconds. The maximum length of native video clips, up from 10 seconds in the previous version. $0.07. The price per second of 720p video, 30% less than Gemini Omni 1.1 Flash. Why it matters: AI video generation is moving from a demo toy to a tool whose cost and duration now fit a real production workflow.
What changes: 15 seconds, conversational editing, fidelity
Gemini Omni 1.2 Flash clears three hurdles at once. First, duration: native clips go from 10 to 15 seconds. Second, editing: the model introduces multi-turn conversational editing — you no longer regenerate an entire clip to fix a detail; you change it with a follow-up instruction, the way you would edit text. Third, fidelity: stronger adherence to image and video references keeps products, characters, and brand details consistent from shot to shot.
That last point is what professionals care about. The nightmare of generative video — a character whose face or logo shifts between two shots — is exactly what reference adherence aims to solve. Google explicitly positions the model as “the next step towards models that can create and edit anything from any input — starting with video.”
The model combines Gemini intelligence with DeepMind’s generative media models. In practice, it accepts short video references to guide a 15-second generation, while first-and-last-frame interpolation stays capped at 10 seconds.
The price drops a third — and the workflow changes
The headline number is $0.07 per second of 720p video, 30% less than Gemini Omni 1.1 Flash, with comparable savings at other resolutions. Taken alone, the price is just a billing line. Combined with the new workflow, it changes how you produce.
The logic Google is pushing is draft fast, then refine: generate quickly at 360p or 720p, fix a detail with a follow-up edit instead of regenerating the whole thing, then upscale to 1080p or 4K. The API produces native 360p or 720p drafts, and the 1080p and 4K options are upscales of those drafts.
The economic benefit is twofold. First, the unit price drops. Second, the number of generations per clip drops: fixing a shot with a follow-up instruction costs less than regenerating the sequence, and a low-resolution draft costs a fraction of the final output. For a studio iterating dozens of times on a shot, the cumulative savings are significant.
The limits to know before you sign
Two limits temper the enthusiasm. The first is the real interpolation length: generation from a short video reference does reach 15 seconds, but first-and-last-frame interpolation remains at 10 seconds. The second is the quality chain: 1080p and 4K are upscales, not native outputs. On a demanding shot — fine text, hair, textures — upscaling does not erase the limits of the underlying draft.
The product also has to be placed in its market. Generative video is a crowded field, and 15 seconds is still short next to players announcing longer clips. The advantage of Gemini Omni 1.2 Flash is not maximum duration: it is the combination of duration, conversational editing, and price — a balance no competitor yet matches on the same API at the same rate.
What it changes for video producers
For a team producing advertising video, product teasers, or social content, the calculus shifts. Reference adherence finally makes generative video credible on campaigns where brand consistency is non-negotiable. Conversational editing moves correction from “redo the clip” to “adjust the shot,” pulling the tool closer to an editing suite than a single-use generator.
The price lowers the experimentation threshold. At $0.07 per second, a 15-second clip at 720p costs $1.05 before upscaling — low enough that concept tests are no longer an expense to justify, but a line of operating cost. That is the change of nature that matters: generative video moves from “pilot project” to “integrated into the flow.”
The deeper issue, common to all generative AI, remains: provenance. As synthetic video becomes indistinguishable and cheap, the question of authenticity — watermarking, provenance metadata, detection — stops being theoretical. Google builds in provenance mechanisms, but the burden of traceability will increasingly fall on producers, not platforms.
Verdict
Gemini Omni 1.2 Flash is not a cosmetic iteration: it is the moment Google’s generative video becomes economically integrable into a production workflow, through the combination of extended duration, conversational editing, and a price cut of a third.
If you produce brand video at volume, test reference adherence on your own assets — that is the capability that decides whether the model enters your pipeline. If you iterate on short shots, adopt the 360p/720p draft, then follow-up edit, then upscale workflow — that is what turns the price cut into real savings. If you need long shots or native 4K quality, wait — interpolation length and upscaling remain limits, not options. One rule to remember: the cost of generative video just stopped being the argument — differentiation will now be won on fidelity and provenance.