Gemini Omni Flash
Google's video model that edits by conversation, in EverAsk.
Try Omni Flash in EverAskRemix the take by asking
Hand it a video and say what to change: a new season, a new style, a new place. It re-renders the same shot, with the same rider, the same bike and the same camera move.
More control, fewer reshoots
Point it at a reference, change a finished take, and trust the physics and the sound to hold.
References drive the shot
Start from a still of a character, a product or a place, up to five images, and the take keeps their look from the first frame to the last.
Change the light, keep the shot
Re-render a take at night, in another season or another style. Drag across to see the same shot, before and after.
Physics that holds
Dominoes fall in order, steel rolls and rings, spray and dust behave the way they would on a real set.
Sound in the same pass
Sizzle, flames, the kitchen around it: the audio is made with the picture and timed to it. Turn it on.
Example takes made on ImagineArt.
From a still to a take in three steps
- 01
Add a reference
Start from a still: a character, a product or a place.
The cat shakes its head like a rapper, to the rhythm of a rap song.
02Write the shot
Say what moves, how the camera follows and what you hear.
- 03
Generate with Gemini Omni Flash
The take comes back with its own sound. Ask for changes, or keep it.
Create from anything
Characters, products, places and pure imagination, all as video.
Example takes made on ImagineArt.
Omni Flash at a glance: six things to know
Gemini Omni Flash is Google DeepMind's video model. Write a shot, start from up to five reference images, or hand it a take to change, and it films a clip of 3 to 10 seconds with its own sound. Then ask for changes in plain words and it re-renders the same shot. In EverAsk it sits beside the other video models, so you can run one prompt on each and keep the best take.
- Made by Google DeepMind, in EverAsk on your plan
- 3-10s
- Single takes from 3 up to 10 seconds
- 720p
- HD at 24 frames a second, ready to post or cut
- Sound
- Effects, ambience, music and voices, made with the picture
- 16:9 · 9:16
- Widescreen and vertical from one prompt
- 5 refs
- Up to five reference images, or a take to edit
Frequently asked questions
What is Gemini Omni Flash?
Gemini Omni Flash is Google DeepMind's video model. It makes video from a text prompt or reference images, with sound, and it edits video when you describe a change in plain words. It is one of the video models in EverAsk.
How does editing by conversation work?
Generate a take, then ask for a change, such as a snowy evening or an anime style. Omni Flash re-renders the same shot with the change, keeping the subject and the camera move.
How long can an Omni Flash video be?
A single take runs from 3 up to 10 seconds at 720p. For something longer, sequence several takes on EverAsk's timeline with music and captions.
Does Omni Flash generate sound?
Yes. Audio is made with the picture: effects, ambience, music and speech. Keep it, mix it under music, or replace it on the timeline.
Do I need a Google account or a Gemini plan?
No. Omni Flash runs on your EverAsk plan alongside the other video models, on one bill.
What shapes can it make?
Widescreen 16:9 and vertical 9:16 for Reels, TikTok and Shorts.
EverAsk on your phone and your Mac
EverAsk for iPhone
Run the same shot on every video model on the train, and pick the winner on your phone. Every clip you make here is on your phone.
Download on the App Store
EverAsk for Mac
Every video model on your desktop, with the stills and footage you already have to hand.
Download for macOS