Google's Gemini Omni 1.1 Flash Brings 4K Upscaling to Video Production

Google shipped an upgrade to its generative video model that pushes it closer to something a production team can actually use. Gemini Omni 1.1 Flash arrives as a production-ready release with scene extension, keyframe control, cheap 360p drafts, 4K upscaling, and video references, all reachable from the Gemini API and Google AI Studio.

The original Omni Flash, released earlier this summer, was where Gemini's multimodal reasoning met video generation and editing. Pricing came in at $0.10 per second of output, matching Veo 3.1 Fast, but the model was stuck on short clips with limited controls. Version 1.1 fills in the missing production knobs.

What actually changed

The headline feature is longer, more coherent scenes. Scene extension takes an existing video and continues generating footage seamlessly from where it left off. The important detail is context: the model can now analyze up to 10 seconds of prior footage before continuing, up from just the final second in earlier versions. You can chain extensions in 10-second increments up to a cumulative 40 seconds.

The rest of the new controls stack up like this:

• First and last frame interpolation. Provide two keyframes and Omni fills in continuous motion between them, useful for camera orbits, zoom transitions, and seamless loops.
• 360p draft mode. Preview renders come back up to 60% faster and cost about a third of a 720p generation, aimed at storyboard iteration.
• 4K upscaling. Final outputs can be pushed to 1080p or 4K for delivery.
• Video references. You can feed up to three seconds of reference video into a prompt to lock in character, style, or motion, on top of the existing text and image inputs.

Calling the model

The model ID is gemini-omni-1.1-flash and it plugs into the same Interactions API used by the earlier preview. Chaining a scene extension looks about as minimal as you would hope:

from google import genai
client = genai.Client()
interaction = client.interactions.create(
    model="gemini-omni-1.1-flash",
    previous_interaction_id=previous_video_interaction.id,
    input=[{"type": "text", "text": "Continue the scene."}],
    response_format={"resolution": "360p"},
)

Notice the previous_interaction_id. That tells Omni to pull the last 10 seconds of context instead of treating your extension as a fresh generation, which is how continuity gets maintained across chained calls.

Where it fits, and where it breaks

Omni's strengths are fairly specific. It leans on an understanding of physics combined with Gemini's knowledge of history, science, and cultural context, and it holds character identity across turns. Subjects introduced in one shot retain their face, clothing, and voice across cuts and subsequent edits in the same conversation, without re-uploading the reference each turn. That behavior makes it interesting for ad creative, product videos, and explainer content where the same subject has to survive multiple edits.

Limits remain. A 40-second cumulative cap beats the original 10-second ceiling but comes nowhere near long-form. Audio input is not supported, so you cannot feed audio clips to synchronize movement to a voiceover. And the model is still a public preview, so pricing, quotas, and regional availability can shift.

Who is already using it

Google name-checked a few partners running the model in production. Adobe wired it into Firefly for video editing. Runway added it to its existing prompt-image-video workflow. Figma's Creative Director on the Weave team, Itay Schiff, called it one of the strongest video models available in Figma Weave and said the 1.1 upgrades take teams beyond generating videos toward directing them. GMI Cloud flagged accuracy for educational content as the reason they shipped it.

The practical read

The interesting shift is that these features line up with a real production loop: cheap drafts in 360p, iteration with keyframes and video references, extension of the winning shot, upscaling to 4K for delivery. Film and ad teams already run that workflow, and now it collapses into one API. For anyone building creative tools, video editors, or automated ad generation on top of the Gemini stack, the Omni documentation and cookbook are the fastest entry points. The model is live in AI Studio today, and Google AI Plus, Pro, and Ultra subscribers get scene extension in the Gemini app and Google Flow as well.