Google has introduced Gemini Omni 1.1 Flash, a new generative video model built around something creators have been asking for since AI video first became widely available: more control. Instead of simply typing a prompt and hoping the result lands close to the idea in your head, Google is positioning Omni 1.1 Flash as a tool for extending scenes, defining shots, using reference footage, generating quick drafts, and turning selected clips into higher-resolution output.

That shift matters. The most impressive AI-generated video clips can still feel like one-off demos – visually striking for a few seconds, but difficult to shape into something that fits a larger story, product campaign, social video, or editing workflow. Google’s latest update is an attempt to move beyond that phase and make AI video more useful for people actually building creative tools and publishing finished work.

Gemini Omni 1.1 Flash is rolling out through Google AI Studio and the Gemini API, while Google AI Plus, Pro, and Ultra subscribers can also access it in Google Flow. Scene extension is additionally arriving in the Gemini app for those subscriber tiers.

The big idea: keep the shot going

The headline feature is scene extension. With Gemini Omni 1.1 Flash, developers can give the model an existing video and ask it to continue what happens next. That may sound simple, but continuity is one of the hardest problems in generative video: a character’s clothes can change, a room can subtly rearrange itself, or a camera move can lose the sense of direction established in the original clip.

Google says Omni 1.1 Flash can analyze up to 10 seconds of previous video context, rather than relying only on the final second of footage. The model can then extend a clip in 10-second chunks up to a cumulative 40 seconds, giving creators more room to construct a sequence instead of settling for an isolated moment.

In practical terms, this could be useful for a short dramatic scene that needs another line of dialogue, an exterior shot that needs a longer camera pullback, or a product visual that needs to evolve from one setting into another. It also opens the door to branching workflows, where creators generate several possible continuations from the same initial scene and decide later which direction works best.

That does not mean the model has solved every continuity problem in AI video. Generative systems can still make surprising choices, especially in shots with complex motion, multiple people, readable text, or specific branded objects. But longer context gives the model a much better starting point than treating each generation as a fresh, disconnected prompt.

More like directing, less like gambling

Google is also adding first-and-last-frame interpolation. Users can provide a starting frame and an ending frame, then ask Gemini Omni 1.1 Flash to generate the video that connects them.

This is one of those features that sounds technical but is easy to understand if you have ever edited video. Imagine having a wide shot of a musician on stage and a final frame that appears to be inside a television screen. Instead of manually creating every movement between those two points, a creator could ask the model to invent a continuous zoom, pan, orbit, or transition that gets from A to B.

The appeal is obvious for ads, social clips, concept videos, motion graphics, real-estate visuals, and experimental editing. A first-and-last-frame workflow gives creators a way to establish the visual destination before asking AI to fill in the journey. Google specifically highlights camera orbits, zoom transitions, and looping clips as potential uses.

That is also a much more sensible creative model than expecting a single text prompt to carry every decision. Prompting will remain important, but keyframes give users anchors. In video work, anchors are often what separate a happy accident from a repeatable result.

Draft fast, then finish properly

Another major part of the announcement is a new 360p draft mode. Google says videos generated at 360p can be created up to 60% faster and at roughly one-third the cost of standard 720p output, based on system throughput.

That may be the most practical improvement in the release. Professional creators rarely produce one version of an idea and call it done. They test pacing, framing, tone, characters, camera movement, lighting, and prompts. Rendering every experimental version at full quality is slow and expensive, especially when most of those early attempts will never make it into a finished project.

A lightweight preview mode allows an editor or creative team to test the core idea first. Does the shot work? Is the camera move too busy? Does the product remain recognizable? Is the transition convincing? If the answer is no, it is better to find out on a cheaper 360p draft than after waiting for several high-resolution generations.

Once a direction is working, Gemini Omni 1.1 Flash can produce 1080p or 4K output, according to Google. The company is clearly trying to make the model fit into a familiar creative pipeline: rough draft, review, revision, final delivery.

Reference video is becoming essential

Text-to-video generation is exciting, but text alone is not always enough. A prompt can explain what should happen, yet it may struggle to convey exactly how a person moves, how a camera swings through a space, or what rhythm a scene should have.

Gemini Omni 1.1 Flash can use up to three seconds of video as a reference in a multimodal input. Google says the feature is designed to help maintain visual context and character consistency while letting users borrow motion or performance cues from short source clips.

For creators, this could mean using a dance reference to guide an animated character, supplying existing footage to preserve the feel of a scene, or combining imagery and reference clips to create a more specific result than text prompts can deliver by themselves. It is another indication that the future of AI video is likely to be multimodal. The strongest workflows will not start with a blank prompt box; they will blend text, images, clips, edits, and human direction.

There are obvious legal and ethical considerations, too. Reference-driven generation can be creatively powerful, but it also raises questions around consent, ownership, recognizable performances, and how closely an output imitates a real person or original work. Those concerns are not unique to Google, and they will continue to shape how platforms build guardrails around AI media tools.

Google wants developers in the middle

This is not just a consumer-facing Gemini update. Google is framing Omni 1.1 Flash as a building block for developers creating their own video-generation products, editing software, creative agents, and media tools. It is available through the Gemini API in Google AI Studio, as well as through the Gemini Enterprise Agent Platform.

That developer-first approach matters because many people will not interact with the model directly. Instead, they may encounter it inside a design app, a marketing platform, a filmmaking tool, or a video editor that wraps the underlying model in a more approachable interface.

Google says Adobe has integrated Gemini Omni Flash into Adobe Firefly, while Figma Weave, GMI Cloud, and Runway are among the companies highlighted as early users or partners. Those names show where Google sees the opportunity: not merely in generating clips for fun, but in becoming one of the engines behind professional creative workflows.

The company is entering a fiercely competitive market. AI video has become a strategic battleground, with models competing on realism, character consistency, prompt adherence, speed, audio, editing controls, and cost. A model that produces a beautiful five-second clip may still lose out if it cannot reliably support revision and iteration. Google’s focus on extensions, keyframes, lower-cost drafts, and upscaling suggests it understands that reality.

What it means for creators

For independent creators, publishers, marketers, and small production teams, Gemini Omni 1.1 Flash could make certain kinds of visual production more accessible. A tech site might use it to prototype a concept trailer for a feature story. A small business could build short product visuals without booking a full shoot for every variation. A video editor could test transition ideas or generate supplementary atmospheric shots.

But the real value will depend on consistency. The novelty of AI video has never been in doubt. The test now is whether a creator can get close to the same character, scene, visual language, and camera intent across multiple clips – and do it without burning through time or budget.

Google’s new release does not make traditional production skills irrelevant. If anything, it puts more value on them. Knowing how to describe a camera move, choose a strong opening and closing frame, judge continuity, edit rhythm, and recognize when an AI generation is not good enough will matter just as much as the ability to write a clever prompt.

Gemini Omni 1.1 Flash is therefore less interesting as a flashy new model name and more interesting as a sign of where AI video is going. The industry is moving away from “look what the model made” and toward “here is how you can shape it.” If Google can deliver reliable control at speed and scale, that may prove far more important than any single viral clip.