In its big keynote at I/O 2026, Google announced Gemini Omni, an AI model that can generate videos from almost anything. In the weeks since this announcement, paid subscribers have been able to combine text, images, documents, and video clips to generate new AI videos. Google is currently rolling out a new Omni feature that lets you generate a hyper-realistic avatar of yourself. Although it’s limited to videos at the moment (there’s no AI profile picture yet), it does a surprisingly good job of creating talking head videos with just a few reference selfies, surpassing anything the short-lived Sora app accomplished. Google has launched a tool with Omni that lets you create your own deepfake videos (and only yourself) in just a few minutes. Are you ready?
Create your own deepfakes with Gemini Omni today
Credit: Khamosh Pathak
Thankfully, you can’t turn your selfies into deepfake videos. You will need to go through a verification process from Google that requires you to scan your face. Once you have access to this Omni feature, go to the Gemini sidebar and movie. Here you will see a pop-up to create your own avatar. If it doesn’t appear, plus Select the button from the Gemini prompt box and tap avatar. Gemini will direct you to a Google site to scan your face. Here, take a few selfies and move your face from side to side. Then just say a few numbers out loud and you’re done. You don’t even have to speak words or sentences. Once the process is complete, your avatar is ready and you can return to Gemini.
Then go back to the Video section and type @.[your name]Use “” to use your own avatar in your video. The video will take a few minutes to generate and you will receive a notification when it is ready. You can play the video in the app, save it to your gallery, or share it via a link or the video file itself. There are some more limitations here. Gemini’s new avatar feature is only available to AI Pro or AI Ultra subscribers using a personal account. What’s more, this feature is limited to users over 18 and in the European Economic Area, Switzerland, or the United Kingdom.Currently, each generation can only produce audio with an obvious Gemini watermark, but Google’s new industry standard for subtly watermarking AI videos and images, SynthID, is also coded in for subtly watermarking AI videos and images with metadata that can be tracked even if the video is cropped.
Would you like to deepfake yourself?
After playing around with this feature all day and generating a series of videos, I found this tool to be surprisingly good. We are not prepared for a time when deepfake technology is so widespread and easily accessible. You can see this in the example below. I asked Gemini to generate a video reviewing the iPhone 17 Pro and request that it include footage of me saying certain sentences. It’s done.
What do you think so far?
Of course, there are limits and there are many drawbacks. I asked you to review the iPhone 17 Pro, but I used the iPhone 16 Pro in the video. At one point, something spontaneously appears on top of one of the many headphones in the background. And although the audio sounds like my real voice, it has no rhythm. No personality, just a flat delivery. And that extends to the video itself. As a human, I’m saying something, but this video is completely lifeless. Too pretty, too sharp, my hair isn’t always beautiful that good. If you know what to look for, it’s not necessarily hard to determine that this is ultimately an AI deepfake (if Gemini’s watermark and YouTube’s AI label didn’t reveal it). But the question is how long that will last. And who would look deep enough?
Currently, the video length is limited to 10 seconds and cannot be edited or modified, so the potential for exploitation here is still very small. However, Google said in a statement that it is currently collecting feedback. The company says it’s “still experimenting when it comes to editing videos and changing audio and audio, and we’re working to better understand how we can responsibly bring this functionality to our users,” so an interactive editing interface could be coming in the future. In the near future, users may be able to show themselves in videos that can be tailored to exact specifications before being sent out to the broader Internet. We may not be ready for that.
