Google has rolled out upgrades to its Gemini Omni 1.1 Flash video model that expand what the system can do while slashing costs and generation time. The model now analyzes up to ten seconds of existing video footage instead of just one second, allowing for more coherent scene extensions that maintain visual consistency across longer sequences.
The core improvement addresses a limitation that plagued earlier video AI tools. When extending scenes, the model previously relied on only the final frame to predict what comes next, often resulting in jarring transitions or characters that drift out of frame unexpectedly. By processing ten seconds of context, Gemini Omni 1.1 Flash understands motion patterns, lighting conditions, and object positioning over a longer window. This produces smoother continuations that feel less stitched together.
The new system can extend video content in 10-second increments up to a maximum of 40 seconds per generation. This means users can build longer sequences by chaining multiple extensions together, though the quality and consistency will depend on how well each 40-second chunk matches its predecessor.
Google introduced a 360p draft mode that represents a significant cost optimization for developers and content creators. This lower-resolution option runs up to 60 percent faster than standard generation while costing only a third as much. The trade-off is visual fidelity, but draft mode works well for testing prompts, iterating on ideas, or generating placeholder content before committing to full-resolution output. For studios working on dozens of concepts or running multiple experiment batches, this cost reduction could shift the economics of AI video production.
The changes target both creative professionals and enterprise users who need faster iteration cycles. Filmmakers and ad agencies experimenting with AI can now test ideas cheaper and quicker. The ten-second memory window makes it practical to extend product shots, establish character movements, or add ambient detail without losing continuity. Content creators working on tight budgets benefit from the lower-cost draft rendering before committing resources to premium output.
This update positions Gemini Omni 1.1 Flash directly against competing video generation platforms like OpenAI's Sora and Runway Gen-3. Sora launched with longer generation windows but higher latency and cost. Runway Gen-3 offers comparable speeds with different memory approaches. Google's emphasis on context window and cost creates a practical middle ground for teams balancing quality, speed, and budget constraints.
The timing reflects broader industry movement toward making AI video generation more accessible beyond research labs and well-funded studios. As generation becomes faster and cheaper, adoption accelerates in marketing, design, education, and entertainment workflows. Google's improvements don't fundamentally change how the model works, but they make it viable for more use cases and more organizations.
These updates arrive as video AI tools mature past novelty phase into production tooling. The focus shifts from "can it generate video" to "can I use this profitably in my workflow." Better context handling and lower costs directly address that question. For teams already experimenting with Gemini video, these changes expand what they can attempt. For teams still evaluating options, Google now offers a more compelling cost-to-quality ratio than before.