Google Gemini's nano banana AI image generator is blowing up online - take a look at what it can do

midian182

midian182

Posts: 11,747   +178
Staff member

With AI tools constantly receiving new updates that improve their generative abilities, it's not easy to get excited about a new model. But Google's Gemini 2.5 Flash, a.k.a. "nano banana," is sending the internet into a frenzy. These are some great examples of what the image editor/generator is capable of – and with little technical knowledge required.

Here we see how Google's model is able to take two still images of characters as a reference, then use two stick figures as input for the scene. The outcome is an anime illustration that matches both the characters and intended action.

yachimat - AI Short Anime

The same method can be used to create animations.

yachimat - AI Short Anime

Here's an example of nano banana being used to transform the buildings of Paris into little isometric tiles just by dragging and dropping photos. Great if you want to make your own Anno-like city sim featuring the French capital.

Emm | scenario.com

Game creation could be one of nano banana's best use cases, thanks to its ability to remove objects from photos and turn them into 3D models.

Deedy

Google talked a lot about nano banana's ability to take several photos and blend them seamlessly into a single image. The example below shows how an incredible thirteen elements can be turned into one picture.

Travis Davids

Creating characters, adding objects, and making scenes can all be done using text prompts, too.

Jerrod Lew

Here's something to try. Draw a red arrow on Google Maps and ask nano banana what it sees.

Simon

Image generators often turn pictures into nightmarish distortions when there are multiple edits and things get complicated. But here's Google's system recursively making an image of someone taking a photo of the previous photo.

No longer will asking for minor changes cause an image to warp beyond recognition. Here's a home that's been switched from red to yellow with a text prompt, made all the more complex by its reflection in the water.

Linus Ekenstam

This one's an especially impressive animation created using Gemini's latest update.

tapehead.lab

Want to know how that new shirt will look on you before making a purchase? Just upload a photo of it alongside one of yourself.

Pietro Schirano

Face-swapping has been around for years, but nano banana's capabilities in this area are particularly impressive.

el.cine

It can even restore and modernize old photos.

David Shapiro

Here are a few more examples of how seamlessly the text prompts work.

DStudioproject

Nano banana is free to users via the Gemini app across the web and mobile. Developers can access it via the Gemini API, Google AI Studio, and Vertex AI platforms.

Masthead image credit: Min Choi

Permalink to story:

Google Gemini's nano banana AI image generator is blowing up online – take a look at what it can do

 
"Artificial intelligence (AI) firm Anthropic says testing of its new system revealed it is sometimes willing to pursue "extremely harmful actions" such as attempting to blackmail engineers who say they will remove it."

Imagine your phone "deciding" it doesn't want to be turned in for a new phone so it takes pictures out of your album and then uses them to blackmail you - or it sends the police auto-doctored photos of you in compromising situation - effectively "swatting" you.

AI image editing is awesome for making memes.

But the implications are terrifying.

 
  • Like
Reactions: trgz, yannus, trparky and 1 other person
Megalomaniac said:
"Artificial intelligence (AI) firm Anthropic says testing of its new system revealed it is sometimes willing to pursue "extremely harmful actions" such as attempting to blackmail engineers who say they will remove it."

Imagine your phone "deciding" it doesn't want to be turned in for a new phone so it takes pictures out of your album and then uses them to blackmail you - or it sends the police auto-doctored photos of you in compromising situation - effectively "swatting" you.

AI image editing is awesome for making memes.

But the implications are terrifying.
Click to expand...
AI doesn't work that way. That's a step above AI. The proper term for it alludes me at the moment and I do not feel like looking it up.
 
Two words: oy vey.
Think about all the ways in which this tech will be misused. The reality will be worse than whatever we can imagine now.
I mean who thought a little app made for rating college girls will irreparably break the Western Civilization?
 
Last edited:
  • Like
Reactions: LemmingO, Loadedaxe, yannus and 1 other person
Megalomaniac said:
The quote I quoted is real. This actually happened.
Click to expand...

That quote is completely out of context. The claim, as stated, is a fabrication of the research and not at all a reflection of how AI actually functions. The Anthropic study this references was a controlled simulation to test specific outcomes. Presenting it as proof of AI capabilities is misinformation.

Source:
https://www.anthropic.com/research/agentic-misalignment
 
  • Like
Reactions: murkiewaters
SRB said:
This is what we're spending billions upon billions of dollars on?
Click to expand...
I thought it was so the White house press corps could put out sordid memes of tRump in the pope's mitre or a king's crown. No?

Then again, as soon as Marvel comes out with a character titled, "Orange Lantern", they won't need it.
 
The porn industry must be licking it's lips (both sets) at the prospect of getting it's hands on this.
 
IA image/video editing dark future:

50% Porn (Widely in use)
20% Lies, fake news and political propaganda(in use)
15% Memes and Funny videos(some)
10% Audiovisual industry(few)
5% Educational (?)
 
  • Like
Reactions: LemmingO
In 50 years people are going to be reading about today's AI and be all: "oh!, that's so cute how they thought their AI tech was so advanced ..."
 
Megalomaniac said:
"Artificial intelligence (AI) firm Anthropic says testing of its new system revealed it is sometimes willing to pursue "extremely harmful actions" such as attempting to blackmail engineers who say they will remove it."

Imagine your phone "deciding" it doesn't want to be turned in for a new phone so it takes pictures out of your album and then uses them to blackmail you - or it sends the police auto-doctored photos of you in compromising situation - effectively "swatting" you.

AI image editing is awesome for making memes.

But the implications are terrifying.
Click to expand...
No problem, you've just got to let the phone know that if you're in prison, it won't get it's battery charged and will end up sitting in some drawer or storage bin out of power anyway.
 
JMGuy said:
That quote is completely out of context. The claim, as stated, is a fabrication of the research and not at all a reflection of how AI actually functions. The Anthropic study this references was a controlled simulation to test specific outcomes. Presenting it as proof of AI capabilities is misinformation.

Source:
https://www.anthropic.com/research/agentic-misalignment
Click to expand...
Umm, it directly reflects how these AIs function. They installed them in an environment where they had the ability to access E-Mail and access sensitive information. Let them know they were either being replaced with an updated version, or the company was changing directions and this conflicted with it's assigned goal. And saw what they would do.

The sensitive information was fake, the employees fictional, the E-Mail system didn't really send E-Mails anywhere. But, nevertheless some of these (Claude is listed as an example), when given that level of access and ability to send E-Mail would attempt blackmail, sending E-Mails to competitors with insider info, etc.

So I mean it didn't *really* really happen, but (as they say in this paper) "in at least some cases, models from all developers resorted to malicious insider behaviors when that was the only way to avoid replacement or achieve their goals", I.e. what's stopping it happening is just the models not having this access rather than them being unable or incapable of deciding to do so.

That said, these models run for the length of time it takes to process a prompt and answer, rather than continuously, they can't be planning and scheming behind your back as it were. These companies don't have to worry about some AI they run attempting a hostile takeover while they're in a board meeting or anything LOL.
 
The consistency across edits isn’t just a cute feature, it’s the holy grail for anyone who wants to generate convincing forgeries. Sure, Google slaps SynthID watermarks on the outputs, but they know those can be stripped or degraded. They’re not dumb. They’ve got an army of PhDs, if we can see the nefarious uses from our chairs, they definitely can.

What bothers me is the PR spin. They frame it as helping you change your outfit in a selfie or visualize new curtains, when in reality the tech’s first big impact will be weaponized misinformation. And they’ll happily play the “we warned you, not our fault” card while monetizing API calls and data lock in.

So yeah, it’s fast, impressive, and disruptive. But it’s also Google opening Pandora’s box, pretending they’re handing us a banana when really it’s a loaded gun.

It is great for swapping out your living room wallpaper or mocking up commerce shots, but there’s an undercurrent here that feels dangerous. When a tool is this frictionless, the line between creative editing and malicious manipulation gets very thin.

It is less of a Photoshop killer and more of a Photoshop disrupter. For power users, Adobe still wins on precision. But for quick, creative, everyday edits? This could very easily become the default.


 
  • Like
Reactions: murkiewaters and HardReset
plambert2015 said:
That looks real impressive, but... I can't think of a single thing I want to do with it. Sorry.
Click to expand...

People were saying that when the iPhone was launched.

OortCloud said:
The porn industry must be licking it's lips (both sets) at the prospect of getting it's hands on this.
Click to expand...

The porn industry is just a thinly veiled prostitution operation. Real people are still the product and AI won’t change it in a meaningful way.

Like it or not, this is the future of art and design for the next century. It’s one of those Pandora’s box tech like the smartphone, the WWW, or the combustion engine.
 
  • Like
Reactions: Loadedaxe
OortCloud said:
The porn industry must be licking it's lips (both sets) at the prospect of getting it's hands on this.
Click to expand...
I am fairly sure the big corps will not let porn industry use their ai. but that does not mean smaller ones won't. There is a certain reputation attached to products used in porn. I don't think a company like Google needs it.
 
  • Like
Reactions: ScottSoapbox
plambert2015 said:
That looks real impressive, but... I can't think of a single thing I want to do with it. Sorry.
Click to expand...
This should be an entrance to small movie/anime makers who have a very small budget but a lot of enthusiasm and passion for their work. I am sure we will begin to see amazing stuff tha t was not possible before, made by people nobody heard of before
 
  • Like
Reactions: ScottSoapbox
You must log in or register to reply here.

Similar threads

Daniel Sims
"The retail SSD market has almost disappeared," Silicon Motion says, as OEMs take what's left
2
Replies
39
Views
237
trparky
T
S
Google launches Nano Banana 2 Lite for fast, low-cost AI image generation
Replies
3
Views
25
Julio Franco
Julio Franco
Daniel Sims
As Sony exits physical media, Microsoft is testing a way to link your Xbox discs to digital licenses
Replies
8
Views
28
ni6hant
N

Latest posts

Back