Maintain your look as you edit
We launched native image editing in the Gemini app earlier this year, and we’ve been working hard to improve it, with particular focus on maintaining a character's likeness from one image to the next. We know that when editing pictures of yourself or people you know well, subtle flaws matter — a depiction that’s "close but not quite the same" doesn’t feel right. That's why our latest update is designed to make photos of your friends, family and even your pets look consistently like themselves, whether you're trying out a 60’s beehive haircut or putting a tutu on your chihuahua.
Just give Gemini a photo to work with, and tell it what you'd like to change to add your unique touch. Gemini lets you combine photos to put yourself in a picture with your pet, change the background of a room to preview new wallpaper or place yourself anywhere in the world you can imagine — all while keeping you, you. Once you're done, you can even upload your edited image back into Gemini to turn your new photo into a fun video.
Bring your vision to life with advanced editing
Here are a few things to try as you explore this new image editing capability:
Give yourself a costume or location change: Upload a photo of a person or pet, and our model will keep their look the same in every image as you place them in new scenarios. Try putting yourself in different outfits or professions, or even see how you’d appear in another decade — all while still looking like you.
Blend photos together: You can now upload multiple photos and blend them together for a brand-new scene. For example, take your photo and another of your dog to create a perfect portrait of you both on the basketball court.
Multi-turn editing: You can keep editing the images Gemini makes — take an empty room, paint the walls, then add a bookshelf, some furniture or a coffee table. Gemini's working with you all along to alter specific parts of an image while preserving the rest.
Design mixing: Apply the style of one image to an object in another. You can take the color and texture of flower petals and apply it to a pair of rainboots, or design a dress using the pattern from a butterfly's wings.
Originally posted https://blog.google/intl/en-mena/product-updates/explore-get-answers/nano-banana-image-editing-in-gemini-just-got-a-major-upgrade/
Description
Comments (62)
... but doctor, tell me: why is my banana so "nano"?
There is "Discussion is turned off for this model."?
Me: "Gemini, give me an image of myself laughing on my throne after I conquer the galaxy"
Google: "here are a few things to try: create a perfect portrait of you both on the basketball court (...) or even see how you’d appear in another decade".
:-/
Either Google is a lot less ambitious than me, or they're making some rather cool assumptions about where I'll "appear in another decade".
On site gen? No thank you.
Contact Google, maybe they'll let you download the weights so you can run it on your H100 cluster.
Come on onsite generation im waiting. I at least want to try this new hype.
It's up now!
lol, begin the mad dash to make it produce BOOBIES!!
The audacity to think anybody cares about corposlop cloud models in a site for local generation...
they don't make money on local gens so.. buckle up buttercup.
You and I may not care, but somebody probably does. I mean, statistically it's probably not zero.
Im pretty sure this is not a site just for local generations, and I would bet that more than half of the accounts use the generator only
@TheP3NGU1N from what im reading they don't make money on cloud gens either lmao
@antonfawkes33350 probably not, least for not for now, but they certainly don't on local gen folks, since those type of folks are probably not likely going to buy buzz/memberships.
@TheP3NGU1N they do make money with pre release models and lora :) if you want there is a way
thanks
Cool
this is crazy as bananas
its free on the gemini app. dont waste your buzz on it here, 60 per generation is extortion
I did the math and it's about double the cost here than it is on Google's AI studio for pro generations on 2k. So $10 for 40 gens here vs $10 for 80 gens there.
I get the feeling that a lot of the uploads in the gallery below don't understand the point of Nano Banana. It has two major selling points. One is that it maintains consistency in character appearance between images. And the other is that it allows you to insert anyone into a SFW image without the need of a Lora.
Though since the latter is banned from Civitai, the showcase should be on the first. So uploading singular images from txt2img just feels like spam.
it feels more like a waste of a valuable resource specific for certain situations.
Well, yes, but it also has text2img prompt adherence like no other model - see my xenomorph/underpants-trumpet-man/woman lying on grass/sign/dog image, all prompted in a single pass.
very good tip
Generation only? What kind of sick, tasteless joke is this?
You have an H100 Cluster to run it locally?
Would the H100s really help? I kinda doubt that Google would give the weights of their most powerful image gen model to someone else...
It's Google, what would you expect?
april fools in september
Often returning same input images as repeated output. Worked when ask 1 output, but when doing batch (4 outputs, as of multiple repeats) it likes to be lazy, by simply returning back the inputs with minimal edits. Maybe wait for micro banana, as this is only nano.
Welcome to censorship suitable for the middle east.
It's Halal to use.
Texas: Hold my beer
Meanwhile me generating uncensored questionable images from middle east.
The buzz price will only be worth it if the generator allows NSFW content. With censorship, Gemini, which is free, is better.
since i was told it's an API call, im guessing no, for NSFW, as its just using an API to connect to Google's NanoBanana's backend
Can you use this in comfy ui?
Probably the same as the one from Google AI Studio, right?
What I thought was really amazing was that it was possible to create images of characters taken from old videos with rough image quality (for example, footage from old special effects dramas), or from a single photo that was created by trying to make a live-action version of a rough drawing from a setting material collection using ChatGPT, to create photos that were as clear and posed as if they had been revived in the present day, and also to create photos from various angles.
However, it is unfortunate that it is not uncommon for the same image to be generated repeatedly, and that the NSFW judgment is too strict (sometimes pure action scenes, such as a character falling to the ground, were caught). I also felt that it was a little difficult to adjust the lighting and the amount of light. It was easy to use lighting that cast the characters in shadows, or backlighting.
In addition, other tools, such as ChatGPT, may be better for converting rough sketches into photos.
Finally, a model that can create gay horse marriage, and not just straight horse marriage that's labeled gay horse marriage
aspect_ratio setting for standard banana (gemini 2.5) is missing?
The aspect ratios provided are the ones we'll have available for the time being.
@theally I mean users can't choose aspect ratios in the generator and can only generate 1:1 images. No aspect ratios setting in the UI.
Seems the setting is missing? Gemini 2.5 banana supports all kinds of aspect ratios.
@reakaakasky Try Nano Banana Pro - it has an aspect ratio selector. Nano Banana, the original, uses the aspect ratio of any input image - it's that way in the API service we're using. But Pro is much better than the original, so I'd recommend switching!
it's heeeeeere!!!!!!
Generation only? @theally
WTF IS THIS? I didn't buy a H100 cluster for nothing?!?! fix this now!
so, let me get this straight (as i may be misunderstanding this), this, is the actual, Nano Banana Pro model from google? is @theally from inside google? did you make this yourself?
Im asuming im missing something as I would not think that google would make something like this appear on a website such as this, or, are there other "closed-source" models on civitAI idk about, like, lets say, NovelAI 4.5 Full for example?
This is Google's Nano Banana model. We host it via API. Any company, or individual, can use Google's models "inside" their own product or service. And yes, check out my profile - we host many other API-only models; OpenAI's GPT Image 1, Seedream 4, Flux Ultra, etc.
@theally dang, ok then. the only sad part tho is that with civitAI's newest update splitting the credits, as i need (i think) the blue ones to generate on this... i have 653 but that was before the split
@theally So this is not the Pro version? Only the Nano Banana 1 version?
@Secret_bamboo There are two versions on Civitai - the original, and the Pro. When you load it into the Generator you'll see a selector at the top.
@Tundra1996 You can use Blue or Yellow Buzz to generate with this model. Or Green, if you're on Civitai.green.
lol it would appear to be little more than an advertisement for Google and his API service, more than anything else. Not the model. Not anything useful for users seeking to download local checkpoints.
@theally is it Civitai hardware or Google hardware that is used for generating those images? Do you like just forward the prompt to Google servers and wait for the result or still process it locally somehow?
@MV261 It's Google's hardware. Here's the Google Gemini API docs - https://ai.google.dev/gemini-api/docs/nanobanana
If not local, useless.
agreed, not touching it until then
I think this model is just way to big for local, it's only real use is to serve as an interesting benchmark, and you can see we are still quite far away from this in local model (even if it's slowly catching up with qwen and ZIT and a bit chroma... still far away)
It's actual name is Gemini 2.5 Flash Image, not Nano Banana!
Someone go tell Google https://blog.google/technology/ai/nano-banana-pro/