Gemini Omni 1.1 Flash adds video controls, but its API listing still carries a preview label

Google announced Gemini Omni 1.1 Flash on August 27 as an update for developers building generative-video workflows, creative tools, and media-editing software. The company says the model is available through the Gemini API in Google AI Studio and describes the update as production-ready. Its advertised additions are scene extension, first-and-last-frame interpolation, lower-resolution previews, and output upscaling.

The central technical change is longer visual context for extensions. Google says Omni 1.1 can analyze up to 10 seconds of prior video, whereas earlier models referenced only the final second. Developers can extend a scene in 10-second increments to a cumulative 40 seconds, and can specify start and end frames to generate a continuous sequence between keyframes. Google also permits up to three seconds of video reference material as multimodal input.

Google frames 360p as the iteration tier and 1080p or 4K as the delivery tier. It says 360p previews can generate up to 60% faster and at one-third of the cost of the standard 720p output, while final projects can be upscaled to 4K. Those figures are provider throughput and pricing claims, not an independently reproduced measure of visual quality, continuity, or end-to-end production cost.

There is a useful qualification in Google's own documentation. The announcement calls Omni 1.1 production-ready, but the Gemini API model catalog lists Gemini Omni Flash under Preview and identifies the endpoint as gemini-omni-1.1-flash. The same documentation says preview models may be used for production but can have more restrictive rate limits and will be deprecated with at least two weeks' notice. Production-ready creative controls and preview lifecycle status are therefore not the same promise.

For implementers, the improvement is most relevant when a workflow needs iterative control over clips rather than a single text-to-video render. The expanded context and keyframe controls can make shot continuation more controllable, while the preview tier offers a cheaper loop before final rendering. Teams should use the exact published endpoint, test real assets for continuity and latency, and account for the documented preview constraints rather than assuming that a 4K option guarantees a finished production result.