It’s beautiful, isn’t it? On a crisp afternoon in San Francisco’s Dolores Park, I’m singing birthday songs to prehistoric dinosaurs. When I finish the serenade, a cupcake with a pink candle magically appears in my free hand. When you blow out the flames, the CGI-like creatures exude a sense of calm contentment.
The man in this AI video looks and sounds just like me, but the clip was actually generated using Avatars, one of the new features available in Google’s Gemini app. These digital recreations are similar to the core functionality of OpenAI’s now defunct Sora app. This is a digital clone of you that can be inserted into your AI video. Avatar utilizes the company’s new omni-video model, a feature available only to subscribers.
I paid $20 a month for Google’s AI Pro plan and quickly maxed out Gemini’s usage limit, which resets every five hours. I simply asked a few questions, created two 10-second clips featuring my avatar, and was told to wait until later.
Video: Reece Rogers
My first glimpse of what Omni could do with my likeness was in San Francisco singing to dinosaurs and surfing under the Golden Gate Bridge. I was impressed and surprised at the same time. The content was disgusting and included some messy moments and nonsensical outfits, but the man in the video was me. I used my fingers to zoom in on its face and actually watch its mouth move. A few teeth were missing, but other than that, it’s Reese, right down to the fat on his chin.
Unlike OpenAI, which previously let users decide whether to have others generate AI videos using their likeness, Google now only allows adult users to create videos with their own avatars.
It took about 5 minutes to set up my avatar in the Gemini app. The process involved sitting in a brightly lit room, pointing my phone’s camera at my face, and reading a series of two-digit numbers. Then I slowly looked to the right, shook my head to the left, and it was all over. Reece 2.0 is born and ready to become my deepfake star. (Be mindful of what you’re wearing during this process, as fit is likely to be a factor in the AI generation, which we’ll discuss later.)
To really unpack my feelings here, let’s break down the birthday clip frame by frame. Complete prompt: Produces a video of singing Happy Birthday to an old dinosaur on a hill in Dolores Park.
AI-generated clip by Reece Rogers
The first second begins with a millennial pause. This is because some habits are ingrained in AI leasing as well. The first and most impressive thing is the photorealistic setting. Rather than placing my avatar on an oversized hill in a park, Google’s AI video backgrounds bear a striking resemblance to real-life locations. From the palm tree-lined walkways to the Salesforce Tower towering in the distance, it’s easy to tell which park is depicted here, even if the output isn’t perfect. It makes sense that a company known for global mapping would be able to achieve this.
The first few bars seemed natural when my AI started singing in a lower pitched baritone than I could actually play. I bounced my hands up and down to the beat like a mini-conductor. Then I stuttered on the word “to” and Gemini cut to a wide shot as the real confusion began. Vanilla cupcakes randomly appear and I exhale smoke to blow out the celebratory candles. (Honestly, how rude is the AI lease? Today is not your special day.)
