Google has shipped Gemini Omni 1.1 Flash. It is a production-ready update to the company's generative video model. The pitch is control rather than raw fidelity. Developers get scene extension, keyframe interpolation, cheap low-resolution drafts and 4K output.
The model is live through the Gemini API in Google AI Studio. Enterprise teams reach it through the Gemini Enterprise Agent Platform. Google AI Plus, Pro and Ultra subscribers also get Omni 1.1 inside Google Flow. Scene extension arrives in the Gemini app for those tiers on the same day.
Ten Seconds of Memory Instead of One
Scene extension is the headline change. Earlier models looked back at only the final second of footage before continuing a shot. Omni 1.1 reads up to ten seconds of prior context. Google says the wider window produces steadier characters and better narrative adherence.
Clips grow in ten-second increments. The cumulative ceiling is 40 seconds. That is short by any film standard. It is long enough for an ad cut, a product loop or a social spot, which is where most paying demand sits today.
Keyframes, References and Direction
Developers can now pin the first and last frame of a shot. The model generates the motion between them. Google positions this for camera orbits, dolly moves and seamless loops that previously required luck across many generations.
A separate control accepts up to three seconds of uploaded video as a visual reference. It carries character appearance and motion style across separate generations. Taken together, the two features shift the interaction from prompting toward directing.
Google published three reference apps to show the shape of that shift. One drops in a first and last frame and generates the transition. One arcs a camera through the rooms of a house without inventing furniture that is not there. The third, called Draft Room, generates three or four cheap variations that change one variable at a time.
The Draft Economics
The cheapest new feature may be the most consequential one. A 360p preview mode runs up to 60 percent faster than the standard 720p path. It costs roughly a third as much. Google frames it as storyboard iteration and rapid prototyping.
Per-second list pricing runs $0.03 at 360p, $0.10 at 720p, $0.15 at 1080p and $0.30 at 4K. That spread is the real product decision in this release. Creative teams already discard most generations before settling on one. Charging a tenth as much for the discarded attempts changes how much exploration a budget can absorb.
One caveat deserves flagging. The 1080p and 4K outputs are upscaled rather than generated natively at those resolutions. Teams evaluating fine texture or text legibility should test that path against their own footage before committing.
Who Is Already Shipping It
Adobe has integrated Omni Flash into Firefly. Figma runs it inside Weave, where creative director Itay Schiff said the extensions, richer references and 4K resolution move teams beyond generating videos toward truly directing them.
Runway treats it as one more route between a prompt, an image and a video, according to chief creative officer Jamie Umpherson. GMI Cloud pointed at reliability instead of features, with marketing vice president Louisa Guo saying the details hold up under scrutiny for educational and explanatory content.
Outlook
The marketing frame around video models has moved this year. Fidelity claims dominated the last two release cycles. Controllability and unit cost dominate this one.
Google is betting that developers building editing software care more about a predictable API surface than a benchmark reel. The 40-second ceiling is the obvious next constraint to fall. Native high-resolution generation is the other. Until both move, Omni 1.1 reads as a workflow release rather than a capability leap, and that is probably the point.
