Google has released Gemini Omni 1.1 Flash, giving developers more control over generative video — the emphasis is controllability rather than raw quality.
Key facts
- Status: Released 2026-08-27
- Capabilities: Scene extension (10s increments up to 40s), first/last frame transitions, 4K output
- How to use it: Use it to bridge shots rather than generate whole videos — pinning both frames is the unlock
What happened
The model adds three capabilities: scene extension, analyzing up to 10 seconds of prior context and extending in 10-second increments to a cumulative 40 seconds; generation of smooth transitions between a specified first and last frame; and 4K output.
What it means for independent sellers
For ecommerce teams making short video, first/last frame control is the practically useful part. The hard limitation of generated video until now has been that you cannot make it start on a frame you chose and end on another, which means the output never slots into an existing storyboard and can only be used as standalone filler. With both ends pinned, AI clips become transitions: from the final frame of a real product shot to the opening frame of a lifestyle scene, with the model filling the gap. That shifts AI video from "generate the whole thing" to "close the seams between shots," and only the second one actually survives contact with a production workflow. The 40-second ceiling is adequate for short-form, but note it accumulates in segments — coherence across a longer piece still needs a human reviewing it.