Google DeepMind just launched Gemini Omni 1.1 Flash, a new suite of creative controls and generative video tools aimed squarely at developers. According to Google DeepMind, the release makes Omni “production-ready for professional use” through the Gemini API in Google AI Studio. This is the moment Omni graduates from a research demo into something you can actually build products on.
The original Gemini Omni brought real-world reasoning to generative creation. Version 1.1 focuses on control and polish, the two things that separate a fun tech demo from software you can ship.
What’s new in Omni 1.1
Google DeepMind highlights a handful of upgrades built for people constructing real workflows:
- Scene extension with real memory. You can take an existing video and keep generating footage from where it left off. The model now analyzes up to 10 seconds of prior context, a big jump from earlier models that only looked at the final second.
- Longer, more consistent stories. That expanded context window means better visual consistency and stronger narrative adherence. Characters, lighting, and motion hold together instead of drifting frame to frame.
- Extend in 10-second increments. You build out video in 10-second blocks, up to a cumulative 40 seconds total. That’s enough room to tell a short story or branch into a new creative direction.
- Faster iteration. Google DeepMind says the updates make generative video more controllable and quicker to iterate on, which matters when you’re testing dozens of variations.
- Built for deployment. The whole point of this release is real-world use: creative tools, media editing software, and generative video pipelines.
Why the memory jump matters
The leap from one second of context to ten seconds sounds small on paper. It isn’t. Most generative video looks great for a few seconds and then falls apart, because the model forgets what it just made. Faces morph. Backgrounds shift. Physics stops making sense.
By referencing a full 10 seconds of prior footage, Omni 1.1 has far more information to stay on track. That’s the difference between a clip you’d post and a clip you’d quietly delete. For anyone building editing tools or longer-form content, consistency is the whole ballgame.
Who it’s for
Google DeepMind is clearly targeting builders, not casual users. If you’re writing software that touches video, these are the people this release speaks to:
- Developers assembling generative video workflows
- Teams building creative or design tools
- Companies shipping media editing software
Access runs through the Gemini API in Google AI Studio, the same front door Google uses for its other developer-facing models.
A note on the limits
The cap worth flagging is length. You can extend video in 10-second steps, but only up to 40 seconds total. That’s a meaningful ceiling if you’re dreaming of full scenes or long sequences. Google DeepMind frames this as a tool for short stories and creative branches, not feature films. Set expectations accordingly.
Where this fits
Generative video is the hottest battleground in AI right now, and control is the new competitive edge. Everyone can produce a flashy five-second clip. The hard part is giving developers predictable, repeatable results they can wire into a product. That’s exactly what Google DeepMind is chasing here.
Moving Omni into a production-ready state on the Gemini API is a signal. Google wants developers building on its stack before rivals lock them in. Expect the length caps to climb and the controls to deepen in future updates, because that’s the direction the whole field is heading.
If you’re building anything video-shaped, Omni 1.1 Flash is worth a look. Full details are available at the original Google DeepMind source.