we_are_coded.by CODE · The world, decoded
БГ
Google DeepMind

Gemini Omni 1.1 looks ten seconds back instead of one, and that is exactly where a scene holds or breaks

Google · event date: 27 August 2026Design

On 27 August Google announced Omni 1.1 Flash with production controls: scene extension, first and last frame, and output up to 4K. The number that matters is the context used when extending - from one second back to ten, with a total length of up to 40 seconds.

In short
  • Scene extension now works from up to 10 seconds of prior footage; previous models referenced only the final second.
  • Added: setting the opening and closing frame, and output in 1080p or upscaled to 4K through the Gemini API in AI Studio.
  • It is also live in Google Flow for AI Plus, Pro and Ultra subscribers; scene extension reaches the Gemini app too.
Checked on28 August 2026Responsible editorTsvetelin IvanovHow we workMethod · Corrections

Anyone who has extended generated video knows the place where it breaks. The eighth second.

Up to there everything holds, then the jacket shifts shade, the hand arrives from the wrong side, the light comes from the left instead of the right. The reason was simple and tedious: the model remembered the last second of the previous clip and continued from it. One second is not context. It is a snapshot.

The facts: on 27 August 2026 Google announced Gemini Omni 1.1 Flash, a set of generative video controls available through the Gemini API in Google AI Studio. When extending a scene the model now analyses up to 10 seconds of prior context, where previous models referenced only the final second; video is extended in 10-second increments up to a cumulative length of 40 seconds. Added controls include specifying the first and last frame, with the model generating continuous motion between the two keyframes, plus 1080p output and upscaling to 4K. Omni 1.1 is also available in Google Flow to Google AI Plus, Pro and Ultra subscribers globally, and scene extension reaches the Gemini app. The announcement is by product managers Anish Nangia and Alisa Fortin. Google quotes feedback from Figma Weave, GMI Cloud and Runway.

Why ten seconds is a different job

Because ten seconds is enough to work out what we are actually looking at: how far the camera sits, where the light falls, with what momentum the person moves, which colour keeps returning. With one second the model guesses from the last frame. With ten it has behaviour to continue.

The other addition - you set the first and last frame and the model finds the path between them - is a small admission of how the work is really done. Nobody in an edit suite thinks in prompts. You think in a start point and an end point, and the movement between them is craft.

The tool is finally asking where we are going, not only what we want to see.

Where the scissors still are

Forty seconds is not a scene in cinema, it is one completed movement at most. A 30-second spot is now within reach - but assembled in ten-second steps, not poured in one go. The difference matters, and we wrote about it three days ago from the other side.

I will try it on our own footage before saying whether it holds. A vendor demo is always built from shots the vendor chose.

If you make short video for clients, look at the first-and-last-frame control first. It is the dullest of the three, and precisely for that reason it will do the most work for you.

The visual is generated code art. No third-party images.
Follow usFacebookLinkedIn
Official primary sources
→Google - Gemini Omni 1.1 Flash lets you build with more control, 27.08.2026
Original: https://wearecoded.com/en/articles/gemini-omni-11-flash-udaljavane-na-scena.html
ShareFacebookXLinkedInTelegramWhatsApp
← Back to all news