Gemini Omni can now extend videos to 40 seconds and upscale them to 4K

Google has updated its generative video model, Gemini Omni 1.1 Flash, to extend videos up to 40 seconds long by breaking them into 10-second segments. The model can now remember up to 10 seconds of previous footage for context when continuing a scene, helping maintain consistency in characters, environments, and overall direction.

Creators have greater control over the beginning and end of their video clips, with the ability to provide both the first and final frames for the model to generate the rest in between. Optional video references allow users to feed Gemini up to three seconds of existing footage for visual context, improving consistency in character appearance across generated scenes.

Developers can now create 360p previews that are reportedly 60% faster than regular 720p output while costing about one-third as much, making it easier to test ideas before investing more time and money into the final production. Omni 1.1 Flash generates 1080p footage or upscales the finished video to 4K, and the model is now available to developers via the Gemini API in Google AI Studio and to businesses through the Google enterprise Agent Platform.

The updated model is also rolling out globally within Google Flow for AI Plus, Pro, and Ultra subscribers, with scene extension accessibility added directly within the Gemini app.