Gemini Omni 1.1 Flash adds scene extension, keyframe control, and 4K upscaling for developers

Google's Gemini Omni 1.1 Flash is now production-ready for developers, adding scene extension, first/last-frame control, and 4K upscaling to its generative video model — plus a cheaper 360p draft mode for iterating before a final render.

What's new

As Google describes it: "Today, we're introducing Gemini Omni 1.1 Flash, a new suite of creative controls and generative video capabilities to support developers," making the model "production-ready for professional use via the Gemini API in Google AI Studio." The headline additions:

• Scene extension: the model can now analyze up to 10 seconds of prior video context before continuing a shot, versus only the final second in previous versions. Developers can chain extensions in 10-second increments up to a cumulative 40 seconds, with Google saying this improves visual consistency and narrative continuity across the extended footage.
• First/last frame control: developers can specify a shot's starting and ending frames and let the model fill in continuous video between them. As Google puts it, "Omni 1.1 generates continuous video between two keyframes, making it ideal for complex camera orbits, zoom transitions, or seamless looping clips."
• 4K upscaling: final output can be upscaled to 4K resolution.
• 360p draft mode: developers can generate lightweight low-resolution previews "up to 60% faster... and at a third of the cost" compared to Omni 1.1's standard 720p, aimed at rapid prototyping and storyboard iteration before committing to a full render.

Omni 1.1 is available now via the Gemini API in Google AI Studio and on the Gemini Enterprise Agent Platform for developers; the underlying model is also rolling out to consumers through Google Flow for AI Plus, Pro, and Ultra subscribers, with scene extension available in the Gemini app for the same subscriber tiers.

Context

Gemini Omni is Google's generative-video model line, positioned by the company as bringing "real-world reasoning" to video generation rather than pure pattern-matching from prompts. This release is explicitly framed as the step that takes the model from a novelty into a production tool: previous versions could generate video but lacked fine-grained continuity controls like keyframe interpolation or long-context scene extension, forcing developers to stitch together shorter clips manually or accept less narrative coherence across a sequence. Google ships developer documentation, a cookbook, and prompting guides alongside the release, and the API takes the new controls directly as request parameters (for example, a previous_interaction_id plus a text prompt to continue a scene).

Why it matters

Generative video tools have generally struggled with two related problems: maintaining visual consistency across a shot that runs longer than a few seconds, and giving developers precise enough control over camera movement and continuity to use the output in real production work rather than as a one-off demo clip. Scene extension and first/last-frame interpolation attack both problems directly, and the cheaper 360p draft tier lowers the cost of iterating on a prompt before paying for a full-resolution render — a workflow closer to how professional video and animation tools already operate. Whether it closes the gap with dedicated video-generation competitors will depend on real-world output quality, but the feature set itself targets exactly the frictions that have kept generative video from being a reliable production tool rather than a demo.