Google ships Gemini Omni 1.1 Flash with 4K output and longer video

Google has released Gemini Omni 1.1 Flash, an update to its generative video model that the company says makes it production-ready for professional use via the Gemini API in Google AI Studio. The release adds scene extension up to 40 seconds, first-and-last-frame control, 360p draft generation and upscaling to 1080p or 4K.

The update arrives as developers race to build generative video into creative tools, media editing software and automated production workflows. Controllability and iteration speed have become the deciding factors for whether these models can move from demos to deployment.

Longer stories through scene extension

Scene extension lets developers take an existing video and continue generating footage from where it left off. The model now analyzes up to 10 seconds of prior context, which Google describes as "a leap from previous models that only referenced the final second."

The company says this improves visual consistency and narrative adherence, allowing developers to build longer stories or branch into new directions. Videos can be extended in 10-second increments up to a total cumulative length of 40 seconds.

Google published a Python snippet showing how developers chain generations through the gemini-omni-1.1-flash model, passing a previous_interaction_id and a text prompt such as "Continue the scene" with a chosen resolution.

Keyframe control for camera work

Developers can now specify the starting and ending frames of a shot. Omni 1.1 generates continuous video between the two keyframes, a feature Google positions as ideal for "complex camera orbits, zoom transitions, or seamless looping clips."

The company demonstrated the capability with prompts describing a whip-pan from a drummer in a beige suit to a saxophonist and ballet dancer, and a zoom into a TV screen revealing the same woman and scene from the beginning of the clip — both specified as "one continuous shot, no jump cuts."

Faster, cheaper drafts at 360p

Omni 1.1 Flash can generate lightweight 360p previews up to 60% faster and at a third of the cost of the model's standard 720p resolution. Google says the speedup is based on system throughput of 360p versus 720p generation. The company pitches this at rapid prototyping, storyboard iteration and quick rendering in developer platforms.

For final output, the model generates polished 1080p or 4K video that Google says is ready for professional production. Sample prompts accompanying the release include a tracking shot of fish, a chipmunk darting across the frame, and a cinematic macro close-up of golden-orange Japanese maple leaves specified as "photorealistic, 4k."

Video references in multimodal input

The model now accepts up to three seconds of video as reference material, letting developers maintain visual context and character consistency across generations. Google's example prompt instructs the model to combine three uploaded dance videos with three character images: a dog performing classical dance, an octopus doing hip hop and a bear breakdancing, all in one continuous shot in a space from a provided image.

Availability and pricing

Omni 1.1 is rolling out across Google's developer ecosystem. Developers can build directly in Google AI Studio, enterprises can access the model through the Gemini Enterprise Agent Platform via the Agent Platform API, and Google has published official documentation, a cookbook and prompting guides covering scene extension, video references and upscaling. The company also published a pricing table for Gemini Omni 1.1 Flash alongside the announcement.

Google says customers are already running Omni Flash in production through the Agent Platform API and showcased videos they created with the model.

Beyond the developer surface, Omni 1.1 is available starting today to all Google AI Plus, Pro and Ultra subscribers globally in Google Flow, with scene extension also available to those subscribers in the Gemini app.

The release signals Google's push to close the gap with rivals in generative video not on raw clip quality alone but on the workflow features — longer context, keyframe precision, tiered resolution and reference control — that determine whether professional teams can ship with these tools at all.