Gemini Omni Flash, Google's new model that turns anything into video

What is Gemini Omni Flash and where did it come from

A year ago, Google released Nano Banana, a model for generating and editing images that helped millions of people restore old photos and turn sketches into finished pictures. Now the company has taken the next step and introduced Gemini Omni, a new family of models that can create any kind of content from any input material, starting with video.

The first model in this family is called Gemini Omni Flash. You can feed it images, audio, video, and text all at once, and it assembles a high-quality video out of all of that, grounded in Gemini's real-world knowledge. From there, you can edit that video through ordinary conversation, just describing in words what you want changed.

The model is already available in the Gemini app, Google Flow, and YouTube Shorts. For now, Omni only creates video, but Google plans to add other output formats down the road, like images and audio.

How to edit video with plain words

The main feature of Gemini Omni is that you can edit video just by describing the changes you want in words, no editing software required. Every new instruction builds on the last one, characters stay recognizable, the physics in the scene doesn't break, and the model "remembers" what happened earlier in the shot.

You can change one specific thing, or transform the whole scene at once. In effect, your video becomes a starting point for something you never could have filmed for real.

Prompt: Make the sculpture out of bubbles.

Why the clips look realistic, physics and world knowledge

Gemini Omni doesn't just draw a picture that looks realistic, the model also reasons about what should happen next in the scene. It combines an intuitive grasp of physics with Gemini's knowledge of history, science, and cultural context, so the result gets closer not just to photorealism but to storytelling that actually makes sense.

For example, the model has a better handle on forces like gravity, kinetic energy, and how liquids behave, which makes scenes more physically believable.

Prompt: A marble rolling fast on a chain reaction style track, continuous smooth shot.

What you can build video from, any references at once

Omni can turn any combination of source material, an image, text, video, or audio, into one cohesive video. For now, only voice references are supported for audio, but Google says it plans to add other audio input types soon.

Prompt: Dynamic sci-fi film style video based on image_0.png. Elements light up similar to video_0.mp4 synchronized to the beat of the music from audio_0.wav

That means you could, for example, give the model a product photo, a screenshot of your brand style, and a camera-motion reference, and get back a video that respects all three sources at once.

Digital avatars and protection against fakes

Google added an Avatars feature that creates a digital version of you using your own voice, so you can generate video that looks and sounds like you. At the same time, the company is still testing broader audio and speech editing capabilities to figure out how to bring that safely to users.

Every video created through Omni carries an invisible SynthID digital watermark. You can verify that a clip was made with Gemini right in the Gemini app, in Gemini for Chrome, and in Google Search.

Gemini Omni Flash vs Veo 3.1, an honest comparison

To be fair, Veo 3.1 still leads on raw cinematic image quality and clip length. But Veo doesn't have the conversational editing that defines Omni, multi-source composition, and generation grounded in real-world knowledge. For iterative editing work, Omni sits in a category of its own, and the two models end up complementing each other more than competing.

If you work with large volumes of video, chances are you'll eventually want both models, just for different jobs.

Who actually benefits from this model

Different types of users get different value out of Omni. Here's who stands to gain the most,

• Content creators and social teams, since the long cycle of redoing a whole clip for one detail shrinks into a short conversation

• Marketers and brand teams, thanks to the ability to combine a product photo, brand style, and motion reference into one video

• Educators and explainer creators, because the model genuinely understands the subject it's depicting, whether that's anatomy or historical events

• Developers, who should plan their video-API architecture ahead of time, since standalone developer access is rolling out in the coming weeks

The value you get from Omni really comes down to how much video you produce and how often you need small fixes rather than full reshoots.

Where the model falls short

Omni has real limits worth knowing upfront. Conversational editing is reliable for roughly four turns, and a single clip caps out at 10 seconds, so longer stories require manually chaining clips together.

A separate challenge is on-screen text in non-Latin languages, Japanese and Chinese characters render unreliably. Content moderation on Omni is stricter than on Veo, and negative instructions, meaning requests not to do something, aren't always followed, so results are worth double-checking manually.

None of these limits are dealbreakers for the tasks Omni was actually built for, but knowing them upfront saves you from wasting time.

In the End

Gemini Omni Flash changes the whole logic of working with video, you no longer need to regenerate an entire clip just to fix one detail, describing the change in words is enough. At the same time, the model has clear limits, short clips and a capped number of reliable edits, worth planning around ahead of time.

The problem is, testing Gemini Omni and comparing it directly against Veo and other video generators isn't always convenient, access is still tied to a Google AI subscription and isn't equally simple to get everywhere in the world.

That's exactly what unitool.ai is for. Get one subscription and unlock access to the most advanced video and image AI models in one place, no foreign card and no extra access headaches. See for yourself how Gemini Omni and other models handle your specific task.