Gemini Omni 1.2 Flash cut per-second pricing 30% to $0.07/s and stretched clips to 15 seconds, but a full-length clip now costs about the same as the old one did.
Vamsi Tallapudi
Manager, Architect Technology at Cognizant
Gemini Omni 1.2 Flash is Google's new video generation model, and it went generally available on November 5, 2026 with a 30% price cut, to $0.07 per second of 720p video, and a longer max clip length of 15 seconds, up from 10. Sounds like a straightforward win. Do the math on a full clip, though, and the "cheaper" model costs almost exactly the same per video as the one it replaced.
That's the part most of the launch coverage skipped. Here's what's actually true, what it costs once you include the fine print, and where the real catch is hiding.
It's Google's latest Flash-tier model for generating and editing short video clips with synced audio, replacing Gemini Omni 1.1 Flash. It's rolling out across the Gemini API, Google AI Studio, Google Cloud console, Flow by Google, Google Vids, Google Ads, and the Gemini app for Google AI subscribers, with the full 15-second clip length reserved for Ultra subscribers.
The model's real upgrades aren't the headline ones. Conversational editing lets you say "make the lighting warmer" or "swap the product color" without regenerating the whole clip from scratch. Reference adherence is noticeably tighter, so a character's face or a product's logo holds up across cuts instead of drifting. Lip-sync and multi-speaker dialogue got better too, and every output still carries a SynthID watermark.
Google bills by resolution, and the spread is wide:
| Resolution | Price per second |
|---|---|
| 360p (draft) | ~$0.02 |
| 720p (draft) | $0.07 |
| 1080p (upscale) | ~$0.10 |
| 4K (upscale) | ~$0.21 |
That $0.07/s figure is the one Google led with, and it's real: 1.1 Flash billed at roughly $0.10/s for the same 720p tier, so the per-second rate dropped 30%, matching the token price drop from $17.50 to $12 per million output tokens. But upscaling triples your cost per second, and a 4K pass on a 15-second clip runs about $3.15 before you've touched the rest of your pipeline. If your workflow upscales by default, that 30% headline discount barely shows up on the invoice.
Here's the catch nobody led with: Omni 1.1 Flash topped out at 10 seconds. Omni 1.2 Flash goes to 15. Do the actual arithmetic for a full-length clip at 720p:
A full-length 1.2 clip costs you five cents more than a full-length 1.1 clip did. The per-second price fell 30%, but clips got 50% longer, and those two numbers almost perfectly cancel out. You're not paying less for video. You're paying about the same for 50% more of it, which is a genuinely good deal if you actually need the extra five seconds, and a wash if you don't.
I'd call this honest marketing with a misleading headline. "30% cheaper" is technically correct and also not the number that determines your bill.
If you're using Flow, Google's consumer creative app, skip the API pricing entirely, because it doesn't apply. A thread on Hacker News flagged that Flow is charging 18 credits for an Omni 1.2 clip versus 10 credits for an Omni 1.1 clip of the same length. That's a credit cost going up, not down, for the consumer product, even while the API rate fell.
Google never claimed Flow pricing tracks API pricing one-to-one, but it's a rough gap for anyone who read "30% cheaper" and assumed it applied everywhere Omni 1.2 shows up. It doesn't. Check the actual credit cost in whichever surface you're using before you plan a content budget around the press release number.
A few things worth knowing if you're building on this:
gemini-omni-1.2-flash, and it does not run on the standard generateContent endpoint like Gemini's text and image models. It's served through the Interactions API, a stateful endpoint where each generation returns an interaction ID you reference in follow-up edit calls. If you've got existing Gemini integration code, this is not a drop-in swap.You don't get much choice. Starting December 3, 2026, any request still pointed at gemini-omni-1.1-flash auto-routes to gemini-omni-1.2-flash and bills at 1.2 rates, no code change required. If your pipeline depends on exact 1.1 output behavior, that's about four weeks from this post to test the new model and confirm nothing in your editing flow breaks.
Worth noting: Runway, a direct Veo competitor, added Omni 1.2 Flash as an option inside its own product on day one. That's either a sign the model is genuinely good enough that competitors want it on their platform too, or a sign that Google would rather have its model running everywhere than win the "which app do I open" fight. Probably both.
The model itself is a real improvement. Better reference adherence and in-conversation editing are the kind of unglamorous fixes that actually matter if you're producing anything beyond a one-off demo clip, and the Arena jump backs that up. What I don't buy is the framing. "30% cheaper" is the sentence every outlet ran with, and it's true in isolation and close to meaningless once you do the 10-second-versus-15-second math. If you were already happy paying about a dollar for a Flash clip, you're still paying about a dollar. You're just getting five extra seconds for it.
If you're deciding between Google's Flash tier and other fast video models right now, it's worth putting this next to how Gemini 4 Carbon handles longer-context generation, and if you want the broader pattern of vendors quietly shifting the real cost somewhere the headline doesn't cover, OpenAI's Images 2.6 pricing is the same story with a different modality. For where Flash-tier models fit against Google's bigger releases this year, see our notes on Gemini 3.8 Flash.
New AI tools, automation workflows, and course drops — straight to your inbox. Join 2,400+ builders.
Jev routes easy judgment calls away from Claude Opus and escalates only uncertain ones, cutting LLM-judge costs over 4x with almost no accuracy loss.
Thinking Machines' Inkling 2 Small: 594B params, 17B active, $0.30/$1 per 1M tokens on Tinker, and Apache 2.0 weights still pending safety review.
Go 1.27's encoding/json/v2 changes 4 defaults that can break your API silently. Here's the migration path, benchmark reality, and when to actually switch.
Comments