Gemini Omni 1.1 Flash

What’s new

• Promoted from preview to GA: the model ID moved from gemini-omni-flash-preview to the production gemini-omni-1.1-flash, released August 27, 2026; Google marked the preview model deprecated as of September 30, 2026.
• Scenes can now be extended: a single generation still tops out at 10 seconds, but scene extension chains clips in 10-second increments up to a cumulative 40 seconds. Google says edits can now draw on up to 10 seconds of prior context, versus earlier versions that only looked at the last second.
• Start and end frames are yours to set: specify the opening and closing frame of a shot and the model fills in the camera move and transition between them.
• A 360p draft mode: generate a lightweight 360p preview up to 60% faster, then upscale to 1080p or 4K once the shot is right.

What it’s good for

• Turning a photo or a line of text into a short video: feed it up to 5 photos plus a text description and it outputs a clip with native audio, no timeline or parameter panel required.
• Fixing one thing in a clip you already made: tell it in chat what to change — swap a character, adjust the lighting, stabilize the shot, change the background — and it leaves everything else alone.
• A scene that needs more than one shot: a single generation runs up to 10 seconds; when that's not enough, scene extension chains on more footage up to a cumulative 40 seconds, with start/end frame control to land clean transitions.

Prompting tips

• Say explicitly when you want one continuous shot: Google's own docs suggest phrases like "in a single continuous shot" or "no scene cuts" — without them, the model may still cut between shots.
• Keep edit prompts simple: Google's guidance is blunt — "Simple prompts work best for video editing. Overly descriptive prompts can lead to unintended changes."
• Pin down what should stay put: add "keep everything else the same" to an edit prompt, straight from Google's own examples.
• Describe sounds by source, not by feeling: "a distant tram, a suitcase latch" works better than "nice sound."
• Write timing in seconds if you want control: the docs accept timecode-style beats like [0–3s] ... [3–6s] ....
• There's no dedicated negative-prompt field — write exclusions into the sentence: Google's docs say negative prompts aren't supported as a parameter, so put them in the prompt itself, e.g. "do not show any text on screen."

Single continuous shot, no cuts. A cup of coffee steams on a kitchen table.
[0–3s] An orange cat jumps onto the chair beside it. [3–6s] It leans in and sniffs the rim of the cup.
Quiet ambience: the cat's paws on the wooden chair, a faint bird call outside the window. No people on screen, no captions.

Known limits

• Consistency and complex motion are still weak points: Google's own model card says maintaining complete consistency through edits, generating scenes with complex motion, and rendering accurate on-screen text all remain a challenge.
• Editing or extending an uploaded video has real limits: the source video can't run longer than 10 seconds, and the feature isn't available yet to users in the EEA, Switzerland or the UK.
• Changing what someone says is deliberately held back: Google says it's still restricting this capability while the team works out how to bring it to users safely and responsibly.

Using it in Nomi

On the generation canvas, add a video card, pick “Gemini Omni 1.1 Flash” in the model picker and choose a mode (Text to video, First and last frame, Multi-reference). Add your references, write the prompt and generate. In the storyboard you can pick it per shot. The first time, connect any one of the providers above under Settings → Models.

Sources

• Gemini Omni 1.1 Flash lets you build with more control (Google blog, 2026-08-27)
• Gemini Omni Flash model card (Google DeepMind)
• Generate and edit videos with Gemini Omni Flash (Gemini API docs)
• Gemini Omni Flash model reference (Gemini API docs)
• Gemini video generation product page (Gemini)
• Introducing Gemini Omni (Google blog, Google I/O 2026)
• Release notes / Changelog (Gemini API docs)
• Gemini Omni 1.1 Flash (Kie.ai integration docs, used to cross-check technical parameters)