Gemini Omni 1.1 Flash Is Now GA: New Video Controls, 4K Upscaling, and a September Migration Deadline

Summary

Google has moved Gemini Omni 1.1 Flash into general availability with a new stable Gemini API model ID, gemini-omni-1.1-flash. The release adds scene extension, first/last-frame interpolation, and explicit resolution control from 360p through 4K.

The bigger change is workflow control. Instead of treating AI video as a one-shot generation task, developers can now draft more cheaply, refine continuity, and reserve higher-resolution output for the stage when a concept is already working.

Quick Take

• Gemini Omni 1.1 Flash is now GA with the stable ID gemini-omni-1.1-flash.
• New controls include scene extension and first/last-frame interpolation.
• Output options are 360p, 720p, 1080p, and 4K; 1080p and 4K use upscaling.
• Google says 360p drafts can be up to 60% faster and cost one-third as much as 720p.
• The preview endpoint is scheduled for deprecation on September 30, 2026.

What changed from preview to GA

Gemini Omni Flash first launched in public preview on June 30, 2026 as a fast model for conversational video generation and editing. The GA release makes that product materially more practical for production workflows.

Why it matters: the useful upgrade is not simply a larger resolution number. It is a clearer draft → refine → deliver workflow for teams building AI-video products.

How the new workflow works

Draft faster in 360p

Google positions 360p as a rapid-prototyping mode. According to the company, 360p video can generate up to 60% faster and at one-third of the cost of the standard 720p resolution.

That creates a sensible workflow for products that generate several candidates before a user chooses one. Storyboards, prompt variations, motion experiments, and side-by-side creative exploration do not always need final-delivery resolution.

Extend scenes with more context

Scene extension lets Omni continue an existing video instead of generating an unrelated next clip.

Google says Omni 1.1 can analyze up to 10 seconds of prior context when extending a scene. Extensions can be generated in 10-second increments up to 40 seconds total.

More context does not guarantee perfect continuity, but it gives the model more information about subjects, composition, motion, and narrative direction than a continuation system that sees only the final moment.

Control the beginning and end

First/last-frame interpolation allows developers to specify two images and generate the motion between them.

That is especially useful for:

• controlled camera transitions,
• product or design reveals,
• before/after sequences,
• architectural movement,
• and creative tools where users want a defined destination rather than an open-ended generation.

Add video references

Google also says developers can reference up to three seconds of video as multimodal input when constructing a new scene. That gives applications another way to preserve motion or stylistic context.

Resolution decision guide

The highest resolution is not automatically the best choice at every stage.

The important wording is upscaled. Google’s release notes explicitly state that 1080p and 4K outputs are produced using upscaling. Calling the release “native 4K generation” would therefore overstate what Google published.

Who should care

The release is most relevant to teams building:

• generative video editors,
• marketing-content platforms,
• storyboarding and previsualization tools,
• design applications with AI-video features,
• creator workflows that compare many variations,
• and media products where users iteratively refine a clip instead of accepting the first result.

For these products, the ability to make cheap drafts and then move toward a final asset can matter more than a single quality claim.

Developer migration checklist

Existing preview integrations now have a concrete migration task.

1. Change the model ID from gemini-omni-flash-preview to gemini-omni-1.1-flash.
2. Review code or UI that assumes 720p is the only working resolution.
3. Add explicit resolution selection where draft-versus-final control benefits users.
4. Consider 360p for high-volume exploration before paying for more polished output.
5. Test scene extension on your own continuity-sensitive workflows rather than assuming every sequence will remain consistent.
6. Describe 1080p and 4K accurately as upscaled outputs in user-facing product copy.
7. Plan to complete the transition before September 30, 2026, when Google says the preview endpoint will be deprecated.

Limits and unanswered questions

GA status does not answer every production question.

Google’s announcement shows how the new controls are intended to work, but developers still need to test consistency, latency, failure rates, and total workflow cost in their own applications.

The company’s “up to 60% faster” 360p figure is also a vendor-reported throughput comparison, not an independent AI World Scope benchmark.

And while longer scene extension should help continuity, more context is not the same as guaranteed character, object, or camera consistency across every generated sequence.

What to watch next

Three signals will matter after launch:

• whether developers actually use 360p as a standard draft stage rather than generating final-quality clips immediately,
• how reliably the 10-second context window preserves continuity across repeated extensions,
• and whether Google expands production documentation around latency, cost, limits, and reference-video behavior.

The preview endpoint’s September 30 deprecation also gives the ecosystem a near-term deadline that should accelerate migration to the stable model ID.

AI World Scope take

Gemini Omni 1.1 Flash is a more interesting release than “Google adds 4K.”

The real upgrade is the workflow around the video model: cheap drafts, longer continuation context, explicit transitions, reference input, and higher-resolution delivery when the direction is already approved.

That makes Omni easier to imagine inside an actual creative product rather than as a standalone demo.

The best way to judge the release will not be by one polished sample clip. It will be whether developers can repeatedly move from idea → draft → refinement → final output with predictable enough quality, speed, and cost to build reliable products around it.