Google's Gemini Omni Model Gets Major Update with Scene Extension and 4K Upscaling

Google has released an update to its Gemini Omni model, a native multimodal video generation and editing tool. The new version, called Gemini Omni 1.1 Flash, allows for more control over video creation and editing.

The key feature of the update is scene extension, which can now analyze up to 10 seconds of prior context when continuing a clip. This is an improvement from the previous model, which only referenced the final second. The new model can generate continuations of up to 40 seconds in total, with some frames edited to make the seam continuous.

The update also includes first/last frame interpolation and video references, which give users more control over camera movement and character consistency. Additionally, draft videos can now be rendered at a lower resolution (360p) for faster processing and reduced costs, with final outputs upscaled to 4K.