Gemini Omni 1.1 Flash Generates and Extends Video

Gemini Omni 1.1 Flash is a video generation model from Google, not a chatbot. Give it a text prompt, an image, or a short video reference, and it generates a video clip.

Google made the model generally available on August 27, 2026, along with creative controls designed for developers. The most useful are scene extension, the ability to specify the first and last frames, and a low-cost draft mode.

If you're looking for a general-purpose assistant, this isn't it. Gemini Omni 1.1 Flash is built to make video, so the key questions are how long its clips can be, what resolutions it supports, how much it costs, and where you can access it.

What Gemini Omni 1.1 Flash Does

Gemini Omni 1.1 Flash turns text, images, and video references into generated video (Google). Google describes the release as production-ready and aimed at developers building on top of it.

The feature Google leads with is scene extension. The model extends videos in 10-second increments up to a total cumulative length of 40 seconds, analyzing up to 10 seconds of prior context each time (Google).

Two more controls shape the output directly. You can specify a first and last frame to steer a transition. You can also supply a video reference of up to 3 seconds, which holds a character or visual style consistent across shots.

• Scene extension: 10-second increments, up to a cumulative 40 seconds, using up to 10 seconds of prior context.
• First and last frame: set both ends of a shot so the model fills the transition between them.
• Video reference input: up to 3 seconds of clip, used to keep a character or look consistent.
• Output resolution: 1080p or 4K, with a 360p draft mode for iteration.

Why the 360p Draft Mode Matters

Google states that 360p draft generation runs up to 60 percent faster and costs a third as much as a full-resolution render (Google). That changes the economics of getting a shot right.

Video generation is an iteration problem before it is a quality problem. A prompt rarely produces the intended shot on the first attempt, and paying full price for every rejected take is what makes generated video expensive in practice.

The workflow the draft mode implies is straightforward. Iterate at 360p until the shot is what you wanted, then render the approved version once at 1080p or 4K.

Where You Can Use It

Gemini Omni 1.1 Flash is reachable through Google AI Studio and the Gemini Enterprise Agent Platform API for developers (Google). Those are the two surfaces for building anything programmatic on top of it.

Two consumer surfaces carry it as well. Google Flow offers it to AI Plus, Pro, and Ultra subscribers, and the Gemini app exposes scene extension to the same subscriber tiers.

Google publishes per-token pricing in a table on the announcement page rather than in prose. Read the current rates on Google's own Gemini Omni 1.1 Flash post before you budget, because generation pricing moves often and a figure quoted elsewhere ages badly.

• Google AI Studio: the developer entry point for testing prompts.
• Gemini Enterprise Agent Platform API: the programmatic surface for building it into an application.
• Google Flow: available to AI Plus, Pro, and Ultra subscribers.
• Gemini app: scene extension for the same Plus, Pro, and Ultra tiers.

The Limits to Plan Around

The 40-second cumulative ceiling is the constraint that shapes what you can make. It comfortably covers a product shot, a social clip, or a title sequence, and it does not cover an explainer or a training video.

Video reference input caps at 3 seconds, so consistency is steered by a short sample rather than a full clip. Expect drift across a longer sequence built from several extensions.

Google does not state a context window or publish benchmark comparisons for this release. Treat quality claims as something to test against your own footage rather than something the announcement settles.

Plan the shot list around those numbers rather than around a script. A 40-second ceiling built from four extensions gives you four decision points, and each one inherits whatever drift the previous step introduced.

• 40 seconds cumulative: the hard ceiling on a single extended sequence.
• 10-second increments: each extension step, using up to 10 seconds of prior context.
• 3-second reference: the maximum video sample for character or style consistency.
• Not stated by Google: context window and benchmark results.

Who It Fits, and Who Should Skip It

It fits teams producing short marketing video at volume, where 40 seconds is the format rather than a limitation. Social ads, product loops, and background footage all sit inside that ceiling.

Across the workflows we have automated for small and mid-sized business teams, generated video earns its place when the alternative is a stock library or nothing at all. It does not replace a shoot when the subject is your actual product or your actual people.

Skip it if you need anything longer than 40 seconds. Skip it if the video must show a real person or a real product accurately. Skip it if your industry requires a synthetic-media disclosure your process cannot yet produce. None of those are fixed by prompting harder.

What Would Change This Answer

A longer cumulative ceiling would open the formats this model cannot serve today, and length is the single limit doing the most to define its use.

A longer video reference input would improve consistency across sequences, which is where multi-extension output is weakest.

Published benchmark results would let buyers compare it to other video models on something firmer than a demo reel. Until Google publishes them, run Gemini Omni 1.1 Flash in AI Studio on one shot you have already produced by hand, then compare the two.