Google has introduced Gemini Omni 1.1 Flash, the stable version of its fast model for generating and editing videos through instructions. The update provides more control over scenes, transitions, and output resolution.
Longer videos through scene extension
One of the main additions is the ability to continue a scene from the point where an existing video ends. The model can examine up to ten seconds of previous footage to better preserve appearance, movement, and continuity.
Each individual output can last between 3 and 10 seconds. Extensions can be repeated in segments until the video reaches a cumulative length of 40 seconds.
Transitions between two images
Gemini Omni 1.1 Flash also lets developers provide a first image and a final image. The model then generates the movement needed to connect both scenes.
This feature can be used for camera movements, zooms, scene changes, or clips designed to begin and end at specific visual points.
Resolutions for drafts and final output
The model supports output at 360p, 720p, 1080p, and 4K. The default resolution is 720p.
Google presents 360p as an option for creating quick drafts before producing the final version. The 1080p and 4K options use upscaling, so they should not be described as video generated natively at 4K resolution.
Available to developers
Gemini Omni 1.1 Flash is available through the Gemini API and Google AI Studio. Google identifies the stable version with the model code gemini-omni-1.1-flash.
The previous preview endpoint, gemini-omni-flash-preview, will be retired on September 30, 2026. Developers still using it will need to migrate to the new stable version.
