Google’s Nano Banana brought an AI photo generator that could create images that resembled real people – and now Google is giving the same treatment to generated video. During Google I/O, the tech giant announced Gemini Omni, a new AI that allows users to generate video from existing videos, photos, or text.
Like Gemini 2.5 Flash Image – better known by its viral code name Nano Banana – Gemini Omni aims to keep the likeness of the person in the inspiration photo or video intact, a task that generative AI has historically had difficulty with.
Gemini Omni – a name that’s just as much of a mouthful as Nano Banana – can generate ten-second video clips, including audio, using existing photos, videos, text prompts, or a combination of those three. Google calls it “Like Nano Banana, but for video.”
The ability to generate videos from existing videos allows users to reimagine footage. Google says that Gemini Omni can transform a character, an object, or the environment “without ever losing the thread of your original scene." That even includes changing the camera angle, as well as the style of the video.
The launch notably comes after OpenAI shuttered its video generator Sora. Google says that Omni will replace its earlier video generator, Veo, inside the Gemini app.
One of the reasons that Google says that Omni is the video version of Nano Banana is that the new AI tool uses avatars. Like with Nano Banana, Google says that digital avatars allow creators to generate videos that still look like themselves.
I tried generating images of myself using Nano Banana when it first came out, and the experiment was both impressive and a bit terrifying. I still haven’t forgotten the images that resulted from that experiment.
The ability to generate videos that look like real people – and have a realistic look to them, as Google says – likely come with some risks. The launch is already raising questions, introducing more “AI slop” in an era where many social media users already complain about AI-generated content.
In a Reddit forum, one user said that making four videos took up an entire Pro Plan usage window. Another commented that “censorship is crazy tight,” suggesting there’s at least some guardrails in place to prevent AI misuse like generating fake political news or graphic content.
But, the slew of AI-related announcements coming out of Google I/O also includes more ways for users to detect if something is AI-generated. The Gemini app already has the ability to detect AI-generated images and videos that use the SynthID watermark, and Google is now rolling out the feature to Search. It will also be added to Chrome “in the coming weeks.”
The Omini-generated videos will come with a SynthID watermark, which means that a check in the Gemini App, Gemini inside Google Chrome, or a check using Google Search will indicate that the video is AI-generated.
You may also like
Ready to create real video? These are the best budget vlogging cameras.
