Google pushed a new video model into developers’ hands on August 27, 2026, and by the next morning it was already reshaping how teams talk about AI video pricing. Gemini Omni 1.1 Flash is now live in Google AI Studio and the Gemini API, and the headline feature isn’t raw video quality. It’s control: the ability to extend a scene, lock a start and end frame, and iterate on a shot the way a video editor would, instead of re-rolling a whole clip and hoping.

Google’s own announcement frames it plainly: “Today, we’re introducing Gemini Omni 1.1 Flash, a new suite of creative controls and generative video capabilities to support developers,” the company wrote in its official post. That framing matters because it signals Google is competing less on “can it make a cool clip” and more on “can a developer actually build a pipeline on top of this.”

What Shipped Today: Gemini Omni 1.1 Flash in Plain Terms

The model’s API name is gemini-omni-1.1-flash, and it sits inside the broader Gemini Omni family that Google has been building out through 2026. Five capabilities anchor the release: scene extension, first and last frame interpolation, video references, resolution control, and conversational editing. None of these are entirely new ideas in AI video, but Google is bundling them into one callable model rather than shipping them as separate research demos.

Scene extension is the feature getting the most attention, and for good reason. The model reads up to 10 seconds of an existing clip before deciding how to continue it, then extends the footage in 10-second increments. Chain enough of those together and a clip can grow to a cumulative 40 seconds. Google says the payoff is “improved visual consistency and narrative adherence, letting you build longer stories or branch into new creative directions,” according to the company’s post.

First and last frame interpolation lets a developer feed in a starting image and an ending image, and the model fills in the motion between them. Combine that with video references, capped at 3 seconds of reference footage, and you get a workflow closer to keyframe animation than one-shot text prompting. Resolution control rounds out the toolkit, letting builders start cheap and only pay for detail once a shot is locked.

The Pricing Model: Cheap Drafts, Expensive Polish

The model ships with four resolution tiers, and the pricing structure tells its own story about how Google expects teams to use it. Draft footage at 360p runs about $0.03 per second, cheap enough to iterate on motion and pacing without burning budget. Google’s default output sits at 720p, and full 1080p and 4K outputs come through upscaling rather than native generation, according to reporting from The Decoder.

ResolutionPrice per secondHow it’s producedTypical use case
360p (draft)$0.03Native generationFast iteration, motion testing
720p (default)$0.10Native generationStandard delivery, social clips
1080p$0.15Upscaled from draftClient review, marketing cuts
4K$0.30Upscaled from draftFinal delivery, broadcast-adjacent work

That draft-then-upscale pattern isn’t unique to Google, but tying it to a single model with a documented per-second price is a meaningful shift for teams trying to budget AI video work instead of guessing at it after the invoice arrives. A 40-second chained extension at 4K, for context, works out to roughly $12 in generation costs alone before any editing or post-production. At 360p draft resolution, the same 40 seconds costs about $1.20, which is exactly why Google is pushing the low-res tier as the default for iteration.

Where Developers Actually Reach It

Google is offering two entry points. Inside Google AI Studio, the model shows up as a no-code playground where anyone can run text-to-video, image-to-video, scene extension, and frame interpolation without touching an SDK. For production workloads, the Gemini API exposes the same functionality programmatically, and Google is also routing enterprise access through the Gemini Enterprise Agent Platform API. The company’s own copy tells developers to “Start building in Google AI Studio,” a nudge toward hands-on experimentation before anyone commits API budget to a pipeline.

A basic scene-extension call through the Gemini API looks like this, based on Google’s published API documentation structure for video tasks:

curl -X POST \
  "https://generativelanguage.googleapis.com/v1/models/gemini-omni-1.1-flash:extend" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "video": {"source_uri": "gs://my-bucket/clip.mp4"},
    "extend_seconds": 10,
    "resolution": "720p"
  }'

Developer commentary picked up on this quickly. One post covering the launch described it as offering “production-ready control,” a phrase that captures what Google is actually selling here: not a novelty generator, but a component builders can slot into an existing content pipeline and call repeatedly with predictable pricing.

From Veo to Omni: How Google Got Here

This release doesn’t come out of nowhere. Google’s video generation effort traces back to Veo 1, announced at I/O 2024 by Demis Hassabis and Douglas Eck, when the model was largely research-grade and gated behind a VideoFX waitlist. It took until October 2025 for Veo 3.1 to arrive as a genuinely productized, tiered release, and Google spent the first quarter of 2026 widening access with a cheaper Lite tier.

What changed over that stretch is context. Earlier Veo-generation models could only look at roughly one second of prior footage when extending a clip. the new model reads ten times that, and Google explicitly frames the jump from 1-second to 10-second context as the core technical upgrade behind this release. By early August 2026, the branding had already started shifting from “Veo” to “Gemini Omni,” folding video generation into Google’s broader multimodal stack rather than treating it as a standalone product line.

DateMilestoneSignificance
May 14, 2024Veo 1 announced at Google I/OGoogle’s first widely discussed text-to-video model, waitlist-only via VideoFX
Oct 15, 2025Veo 3.1 releasedFirst stable, tiered (Lite/Fast/Quality) Veo generation
Mar 31, 2026Veo 3.1 Lite tier completes rolloutBroader, cheaper access for high-volume use
Aug 7, 2026Gemini Omni video branding appearsGoogle folds video generation into the Gemini Omni multimodal family
Aug 27, 2026Gemini Omni 1.1 Flash launchesScene extension, frame interpolation, per-second pricing, Gemini API access

That two-year arc, from a research waitlist to a metered API product with a documented price list, mirrors how Google rolled out large language models a few years earlier. Gate it behind a research team, open it to a narrow developer group, then productize it once the failure modes are understood well enough to put a price on them.

Gemini Omni 1.1 Flash vs. Sora, Runway, and Kling

Google isn’t launching into an empty field. OpenAI’s Sora line, Runway, and Kling have spent the past two years building reputations around specific strengths. Sora leans on longer, temporally coherent clips. Runway leans on editorial and post-production tooling aimed at working filmmakers. Kling leans on high-fidelity output favored in parts of the Asian creator market. None of those companies have published pricing or context-window specs directly comparable to what Google disclosed today, so a spec-for-spec table would mean guessing, which isn’t something worth doing in a piece about a company that just published exact numbers.

What’s clear from the announcement itself is the angle Google picked. Rather than chase maximum clip length or photorealism as the headline metric, Google’s model leads with controllability: keyframes you can specify, scenes you can extend deliberately, and a metered pricing ladder that lets a team draft cheap and pay only when a shot is locked. That’s a pitch aimed squarely at studios and agencies that already run production pipelines and want an API they can slot in, not a single flashy demo clip that happens to go viral once.

The practical effect: teams evaluating AI video vendors right now are less likely to ask “which one looks best” and more likely to ask “which one fits our review-and-approval workflow without blowing the render budget.” Google built pricing tiers and frame-level control specifically to win that second question.

Market Impact: Budgets, Agencies, and the Render Bill

The per-second pricing model changes how agencies and in-house creative teams will plan AI video spend. A studio producing 50 short marketing clips a month at an average of 20 seconds each, drafted at 360p and finished at 1080p, would pay roughly $30 for drafts and $150 for finished output under Google’s published rates, a total well under $200 a month before labor. That math is small enough that the real cost driver shifts from “can we afford this” to “how many iteration cycles does our creative process actually need.”

It also puts pressure on the stock footage and freelance motion-graphics markets, particularly for short-form social and ad content where a 10-to-40-second clip is often the whole deliverable. Not every use case survives that pressure. Anything requiring brand-specific characters, exact product shots, or legal sign-off on likeness still needs a human production pipeline, and Google’s tooling doesn’t change that.

Enterprise buyers get a separate path through the Gemini Enterprise Agent Platform API, which suggests Google is also chasing larger contracts where video generation gets bundled with existing Google Cloud and Workspace spend rather than sold as a standalone line item.

Why Google Is Moving on Video Right Now

Timing matters here. Google I/O 2026 already passed without a major standalone Veo announcement, and the shift to Gemini Omni branding in early August suggests the video team spent the summer folding its work into the same infrastructure as Gemini’s text and image models rather than shipping a splashy new Veo generation on its own. That’s a deliberate bet: instead of a headline-grabbing model launch, Google is optimizing for a single API surface developers already use for everything else in the Gemini family.

There’s also a competitive clock running. Every month Sora, Runway, and Kling spend building developer relationships is a month Google isn’t capturing that integration work. Shipping scene extension and frame interpolation as callable API features, rather than consumer-app-only tricks, is Google’s way of getting into production pipelines before habits calcify around a competitor’s tooling.

Early Developer Reaction

Reaction in the hours after launch centered on the pricing transparency as much as the features themselves. Coverage from StartupHub.ai noted that publishing per-second, per-resolution pricing at launch is unusual enough in this category to be a talking point on its own, since most AI video vendors have historically bundled pricing into opaque credit systems that make cost forecasting difficult. Google posted the launch directly from its own account, describing the release as its “newest multimodal model for video generation and editing” with a feature list matching the official blog post.

The developer-facing framing, treating extend and image-to-video as documented API tasks rather than app-only features, is what’s drawing early interest from teams that build tooling on top of AI models rather than teams that just want to generate a one-off clip.

What’s Still Unconfirmed

A few details remain open as of this writing. Google hasn’t published a separate consumer-facing “Omni” brand distinct from the API model name, so it’s not yet clear whether Omni becomes a standalone product line or stays a model-family label inside Gemini. Rate limits, quota tiers for free versus paid accounts, and any regional availability restrictions weren’t detailed in the initial announcement either. Teams planning to build on this today should check current documentation directly rather than assume terms will stay fixed, since Google has a track record of adjusting API pricing and limits within weeks of a launch.

Five Predictions for What Happens Next

  • Expect a consumer-facing rollout within Gemini app flows. Google already references Omni inside Gemini’s app tools, and a simplified version aimed at non-developers is a logical next step within the next quarter.
  • Competitors will publish clearer pricing. Google forcing a transparent per-second, per-resolution price list puts pressure on rivals still running opaque credit systems to do the same.
  • The 40-second ceiling will get pushed higher. Chained 10-second extensions feel like an intermediate step, and a future Omni revision extending native context further is a reasonable bet given how fast the 1-second-to-10-second jump happened.
  • Enterprise bundling deepens. Expect Google to fold Omni video generation tighter into Workspace and Google Cloud contracts rather than keep it purely pay-as-you-go.
  • Agencies will build middleware, not switch wholesale. Rather than picking one vendor, expect production shops to build routing layers that send drafts to whichever model is cheapest that week and finals to whichever looks best, keeping Google, OpenAI, and the rest all in rotation.

How to Get Started Today

Developers curious about the model can start in Google AI Studio without writing code, testing scene extension and frame interpolation on a small clip before committing to an API integration. For teams planning production use, the sensible path is drafting everything at 360p first, since $0.03 per second makes it cheap to test five or six variations of a shot before spending on a 1080p or 4K render. That single habit, iterate cheap, finalize expensive, is really the design philosophy behind this whole release.

Frequently Asked Questions

What is Gemini Omni 1.1 Flash?

It’s Google’s new multimodal video generation and editing model, launched August 27, 2026, available through Google AI Studio and the Gemini API under the model name gemini-omni-1.1-flash.

How much does Gemini Omni 1.1 Flash cost?

Per-second pricing runs $0.03 at 360p, $0.10 at 720p, $0.15 at 1080p, and $0.30 at 4K, according to reporting from The Decoder. 1080p and 4K outputs are produced via upscaling rather than native generation.

How long can a video be with scene extension?

Each extension step adds up to 10 seconds and reads up to 10 seconds of prior footage as context. Chaining extensions can bring a clip to a cumulative 40 seconds.

Is Gemini Omni 1.1 Flash free to use?

Google hasn’t published a free tier specifically for the new video capabilities in its initial announcement. Developers should check current Gemini API pricing documentation for the latest quota and billing details before building a production workflow.

How is this different from Veo?

Gemini Omni 1.1 Flash folds video generation into Google’s broader Gemini Omni multimodal family rather than keeping it as a standalone Veo-branded product. The core technical jump is context: earlier Veo-era extension read about one second of prior footage, while Omni 1.1 Flash reads up to ten.

Can I specify the start and end of a video clip?

Yes. First and last frame interpolation lets developers supply a starting frame and an ending frame, and the model generates the motion connecting them, similar to keyframe animation.

Does Gemini Omni 1.1 Flash support 4K output natively?

No. 4K and 1080p are produced through upscaling of lower-resolution drafts rather than native generation at those resolutions, based on Google’s documented resolution tiers.

Where can enterprises access the model?

Beyond Google AI Studio and the standard Gemini API, Google is routing enterprise access through the Gemini Enterprise Agent Platform API.