Omni Flash is the first of these models to be released, now available in the company's AI video generation and editing platform, Flow.
Gemini Omni is the next step.
Gemini Omni Flash is available inside the Gemini app, Google Flow and YouTube Shorts.
We put Gemini Omni Flash to the test, creating 10-second video clips with added audio and image references.
All of this is done simply by chatting with the Agent within your project on Google Flow.
Google has a new tool called Omni. It is a new set of generative models that can turn any kind of input into video. Omni Flash is the first of these models to be released, now available in the company's AI video generation and editing platform, Flow. Last year, Nano Banana brought Gemini's intelligence to image generation and editing, and has since evolved through several iterations. Gemini Omni is the next step.
Gemini Omni Flash is available inside the Gemini app, Google Flow and YouTube Shorts. In time, the company will support output modalities like image and audio.
The new model lets you prompt for AI art and videos in a number of ways — text, images, and uploaded video — with the ability to make modifications, edits, or additions along the way.
In Flow, it all revolves around the Agent feature. Speak to the agent within an ongoing project and you can build videos, characters, and images all in one place, referencing and combining them as you go. Once you've set up your agent, you can include pre-context instructions — so if you want all your generations to match a certain product, brand, colour, pattern, edit, or musical style, you can bake these into the instructions from the start for faster, more consistent results.
We put Gemini Omni Flash to the test, creating 10-second video clips with added audio and image references. The goal was to build an ad campaign for an iced Americano brand called Bean Up — targeting young people who might enjoy a can on the beach, by the swimming pool, at a picnic, or simply while chilling at home. The drink contains zero sugar and is made with high-quality Arabica beans. The series of videos was set across Calcutta, Bengaluru, and Mumbai.
The output was reasonable — good enough for a few social media posts and a laugh or two. That said, you'll need a fair number of credits to get through the entire process. We're currently on the Google AI Pro (5TB) plan at ₹1,950 a month. Depending on your needs you can keep paying, but be warned: this can become addictive and expensive.
We also used the new model to create three ten-second videos of a real-world stuffed toy named Stuffy. It looks a bit like the dragon from Doc McStuffins. The idea was to make Stuffy travel from Calcutta to London and then have tea at Buckingham Palace — just the perfect fodder to keep toddlers happy. First, we took a picture of Stuffy and uploaded it into Flow. Then the magic of prompting began.
On the editing side, you can drag in a video you've created or upload one from your device, then prompt the agent to insert new elements, modify things within a scene, or remove elements entirely — such as clearing cars from around a motorbike. You can even swap out the background of a scene while preserving the motion of a character, so if you have footage of someone running through a location, the shots will still match the original movement, adjusted to fit the new setting. All of this is done simply by chatting with the Agent within your project on Google Flow.
One of the biggest new additions is the characters and avatar feature. You can describe how a character looks, then save them as a reusable asset and reference them by name across multiple scenes and prompts. Each character can be assigned a consistent voice, and whenever they appear, their look and voice are preserved — no need to keep re-prompting. You can combine characters too, referencing two in a single prompt to have them interact.
One of the most exciting new elements in Flow is the Tools section, which takes vibe coding to a whole new creative level. You can build any tool you can think of to assist with your video creation — browse what others have already made, such as colour palettes, filters, scene comparison tools, and drawing overlays, or code your own directly within Flow using Gemini. The tool is built in front of you, usable like an app, and publishable once you're happy with it. One we've found particularly useful is a grid tool that lets you upload four to eight clips, organise them into a layout of your choosing, and download the result — perfect for the kind of comparison content that performs well on social media.
