Google DeepMind
Gemini is getting serious about video.
Omni 1.1 Flash reshapes the production flow
Up to 40 seconds of scene extension, first/last-frame camera control, and native 4K upscaling. Google DeepMind shipped its dedicated video model, Gemini Omni 1.1 Flash, on August 27, 2026. The "Omni" family that unifies text, image, audio, and video has entered production.
The Announcement
Google's video model
used to be an also-ran
In the video-generation race, specialist tools like Runway, Luma, and OpenAI's Sora have led the field, while Gemini lagged behind its own strength in image and audio. Precise scene extension and camera control were territory the specialists still owned.
Google DeepMind just moved on that gap. According to its official blog post, the new Gemini Omni 1.1 Flash model was announced on August 27, 2026, rolling out to developers and enterprises through the Gemini API, Google AI Studio, and the Gemini Enterprise Agent Platform. Google AI Plus, Pro, and Ultra subscribers are also gaining access through the video app Google Flow.
| Gemini video, before | Omni 1.1 Flash |
|---|---|
| Scene extension was limited | Extends up to 40s, reading 10s of prior context |
| Camera control was weak | First/last-frame control for continuous shots |
| Drafts cost the same as final output | 360p draft mode: up to 60% faster, ~1/3 the cost |
| High resolution needed a separate step | Native upscaling to 1080p and 4K |
Video generation, moving from a specialist's outpost
to a standard feature of your own stack.
By The Numbers
Omni 1.1 Flash,
by the specs
How It Works
Set the start and end frame,
let the model fill the middle
Instead of scripting every camera move, you hand the model a first and last shot.
Provide two images
Prepare the images for the first and last moments of the shot — a rough sketch or a still both work.
Draft at 360p
Generate a 360p draft at about a third of the final cost, and check composition and camera movement first.
Export natively to 4K
Once a direction is locked, upscale only the final cut to 1080p or 4K on export.
Who Benefits
Who it helps, and how
Designers & video makers
A still from Figma or Canva can serve directly as a start frame. Run cheap 360p drafts across several ideas, then upscale only the final pick to 4K — cutting manual camera-move tuning.
Marketers
Forty seconds fits most vertical social ad slots. Extending scenes without a reshoot means seasonal or campaign variants can be produced without new footage.
Business leads
Video production could now fold into a single Gemini contract, giving leverage to revisit standalone deals with tools like Runway — though switching costs still need weighing.
What's Next
Plenty is still preview —
don't get ahead of it
Why this matters now. Gemini had a presence in image and audio but trailed specialists in video. Folding a dedicated video model into the same Omni family that already spans text, image, and audio moves video generation from a standalone experiment to a standard capability inside Google's own stack.
What to do next. The first practical move is to exploit the cheap 360p draft mode to test several camera-move ideas before committing to a final render. It's also worth benchmarking image quality and camera fidelity directly against incumbents like Runway. Note that only some capabilities, like scene extension, have reached the standard Gemini app so far.
Risks and counterpoints. Some announced features remain in preview or beta, and specialists may still hold an edge on complex physics or heavily choreographed action. Teams already invested in Runway or Luma face a real evaluation cost before switching. Rights around generated output also warrant watching as Google's terms of service evolve.