Google Creative Lab's Alexander Chen teases an experimental Gemini Omni prompt that animates a digital photo into physical space

I input a video with this prompt:

Start with the computer screen content exactly the same as in the video. Make the bird come to life, turning head, leaving the photo and jumping onto the person's hand.

Gemini Omni 🐦 prompt in 🧵

This is a photo of a zebra finch I got to encounter in real life. Such a cool bird!

@alexanderchen Look at the screen reflection! The prompt didn't even mention it, but Omni nailed the physics. Absolutely insane detail. 🤯

@alexanderchen Ok, Flow seems better

@alexanderchen thank you for sharing what a cool idea!

@alexanderchen Another great one @alexanderchen ! Do you use Flow or Gemini app? Does it matter?

@alexanderchen 我玩了一把御剑决

@alexanderchen Love this!

@ManabeT69109 Oh whoa didn't notice that until you pointed it out. Really cool.

@alexanderchen It was in my bucket list.. Similar concept but different subject

@alexanderchen wow that is just awesome

@alexanderchen awesome! every time I try to use a reference that shows a computer screen with an app on it, it gets flagged and won't generate. any idea what's going on there? @alexanderchen

@gcduncombe Ah, not sure unfortunately. I haven't encountered that I will keep an eye out if I hit that issue.

@alexanderchen Are you just using the Gemini app?

@alexanderchen @demishassabis @zachking 👀

Recursive Omni! Fun idea. The actual underlying model should be the same in both Gemini app and Flow. So either is fine for doing one off model tests. You might just see some differences due to model variability. @FlowbyGoogle has frontend that's useful for folks wanting to go a bit deeper on creating.

@Soranlan Nice!

@alexanderchen 😍😍😍

@alexanderchen Hmm strange! Yea I tried a few different things but on my end it seems consistently blocked