Google is a great example of how a big company with a lot of smart engineers can race from behind and regain domination in a market segment: Google’s Gemini is now one of the very best image and photo manipulation tools available. It wasn’t always this way; Google’s various AI efforts started with BERT, then LaMDA and MUM, then PaLM, then Bard, its poorly received AI released to compete with ChatGPT. It was only at the beginning of 2024 that Bard was renamed Gemini, an AI system that keeps evolving at an astonishing speed.
Summer of ’24 Gemini gained the ability to add and remove objects from images, restyle images, and more. Much of this was promptly folded into Google Photos, particularly on the powerful Pixel Android phone lineup. Six months later Google added “nano banana”, an image model designed to be run locally on devices. Today, Nano Banana’s integrated into Gemini on a lot of devices, including your Pixel phone. But not, surprisingly, entirely in the Photos app…
COFFEE SHOP IN GOOGLE PHOTOS
To test things out, I grabbed my own Google Pixel 9 phone and took the following photo of a local Fort Collins, Colorado coffee shop:
Certainly an attractive café, but what would it look like ten years after an apocalyptic event? A perfect (if slightly weird) job for Gemini. It’s already in Photos, so I tapped on “Ask” and prompted: “add jungle vines to the walls”.
No bueno:
As you can see, it turns out that Google Photos has AI features in its photo editing suite, but it’s really just focused on simple manipulations, not more imaginative applications.
SWITCHING TO GOOGLE GEMINI
There’s a separate Gemini app on the Pixel phone, so I launched that and used the paperclip icon to add the photo:
Tapping on the input field, I specified…
Did it complain or refuse? Nope. In fact, it did a quick update to the image that certainly makes the café more visually interesting:
Definitely more of a jungle theme! But what about abandoning the cafe? This is a much more complicated query because it has to make assumptions and inferences about what it would be like in that circumstance. Here’s my prompt:
Did you catch that I misspelled “people” and it still worked? And did you see that it used “Nano Banana Pro”? A bigger pic of the result:
Getting there, but why are the lights on if it’s post-apocalypse? Not to worry, one more prompt does the trick: “turn off the lights, add more dust, dirt, and mess, as if it hadn’t been visited in years”. The result is… well… here ya go:
Pretty darn good for just a few minutes of manipulation with Gemini, I’d say. You can optionally add zombies and roaming wild animals, as desired.
Pro Tip: I’ve been exploring AI since way before the release of large language models and generative AI. Please check out my AI and ChatGPT Help Library while you’re visiting! Thanks!
