Gemini Omni 1.1 Flash Adds 4K Upscaling and Longer Videos

Gemini Omni 1.1 Flash is Google’s upgraded multimodal video generation model, adding longer scene extensions, frame controls, previews, and 4K upscaling.

Google introduced Gemini Omni Flash a few months ago, and it has now followed up with Gemini Omni 1.1 Flash.

Unlike traditional text-to-video models, Omni can work with text, images, audio, and video as inputs.

Gemini Omni 1.1 Flash can extend videos up to 40 seconds

One of the biggest upgrades involves video extension.

Gemini Omni 1.1 Flash can analyze up to 10 seconds of preceding footage before generating the next part of a video. The previous model used only the final second for context.

Google says the larger context window should help maintain consistency across characters, motion, and overall scene direction.

Users can extend videos in 10-second increments, with a maximum total length of 40 seconds.

First-and-last-frame controls offer more precise generation

Google has also added first-and-last-frame controls.

Users can provide one image for the opening frame and another for the final frame. Gemini Omni 1.1 Flash then generates the video sequence connecting the two images.

Gemini Omni 1.1 Flash supports 4K output

The model supports 360p, 720p, 1080p, and 4K output.

Google uses 720p as the default generation resolution. According to its Gemini API documentation, the system upscales 1080p and 4K videos rather than generating them natively at those resolutions.

Gemini Omni 1.1 Flash availability

Gemini Omni 1.1 Flash is available through the Gemini API in Google AI Studio and Google’s Gemini Enterprise Agent Platform.

Google has also made the model available globally in Flow for Google AI Plus, Pro, and Ultra subscribers.

Eligible subscribers can use the new scene extension capabilities through the Gemini app as well.

In other AI news, OpenAI lost control of 1,200 agents before the Hugging Face breach, while Anthropic recently introduced a unified Claude memory that works across Cowork and chat.