Gemini Omni 1.1 Flash shifts Google’s video API toward iterative editing

Google detailed Gemini Omni 1.1 Flash on August 27 as an update for video generation and editing through the Gemini API. The documented additions center on controls that fit a production workflow rather than a single prompt-to-clip request: scene extension, interpolation between supplied first and last frames, 360p previews, and upscaled 1080p or 4K output. Google says the model is available through AI Studio and its enterprise agent platform. Those are concrete interface and workflow changes, but this report does not treat the announcement as a release-state signal; release confirmation remains outside editorial coverage.

The most consequential change is the extension path. Google says Omni 1.1 can use up to 10 seconds of prior context, where earlier models referenced only the final second, and can append footage in 10-second increments to a cumulative 40 seconds. That does not guarantee continuity in every clip, but it gives editors a more explicit mechanism for carrying characters, motion, and audio through a longer sequence. For teams prototyping narrative or product-video tools, the useful comparison is therefore not simply output quality: the update increases how much previous material the model is designed to consider while continuing a scene.

The API documentation places these controls in the Interactions API, which supports iterative edits using a previous interaction identifier instead of requiring a user to describe the whole video again. It also documents text, image, video, and audio-related inputs, configurable aspect ratio, and a resolution setting whose default is 720p; 1080p and 4K are described as upscaled outputs. The public material describes capabilities and integration patterns, not an independent benchmark against competing video systems. Technical users should therefore separate the documented control surface from any broader claim about comparative generation quality.

There are material constraints. Google documents that uploaded videos used for editing or extension are limited to 10 seconds in several cases, extension can only append to the end of a clip, voice editing is unsupported, and editing or extending uploaded videos is unavailable in the EEA, Switzerland, and the United Kingdom. It also notes that reference videos are capped at three clips of up to three seconds and that multi-video reasoning can degrade results. In practice, Omni 1.1 looks best suited to short, stateful creative iterations with planned handoffs, rather than unrestricted editing of an existing long-form production.