Gemini 2.5 Flash Image - Meet "nano-banana🍌" the new SOTA image model
Gemini 2.5 Flash Image (aka "nano-banana") is Google's new SOTA image model. It excels at character consistency, multi-image fusion, and precise editing with natural language. Available via API, Gemini app and in Google AI Studio.
Replies
Flowtica Scribe
Hi everyone!
About a week ago, a mysterious model called "nano-banana"🍌 appeared on LMArena. Everyone was guessing who made it, but the signs pointed to Google. And now it is.
Its character consistency is what shocked the world, but the model is much more than that. It can fuse multiple images, make precise edits with natural language, and it's all powered by Gemini's world knowledge. I can't even remember how many times I tried to "draw the card" on LMArena this past week, just hoping to get a chance to use it.
It's just that good. I think this is another milestone moment for multimodal models, right up there with GPT-4o. But words don't do it justice, you have to try the model for yourself. I think anyone will be as deeply shocked by its capabilities as I have been this past week🤯
PhotoFuse
I have tried it already and it's way ahead of all other competitors! Can't wait to integrate it into PhotoFuse!
Is there a special way to prompt it to get results/modifications? I tried getting it to revise images but it just keeps sending me the same one or changing things I did not ask it to.
Used it by accident before I read the announcement and was impressed with how good the output was. I expected something maybe passable and got 🤯.
Camocopy
It's crazy where we have come from. Google is cooking with it's new gemini-2,.5-flash-image particularly in the field of virtual try-on apps. I just vibe coded a small virtual try on website with it and the results are impressive. It maintains the original details of the t-shirt flawlessly, without any of the undesirable artifacts that often plague similar applications. It's just nice. The overall experience is smooth and visually appealing, I can't think where we will be in 1 year 😅
Also the model itself is nice for vibe coding too, so you can vibe code an app where you will use the same model. Seems paradox. @zaczuo
Tried it on LM Arena and immidiatly fell in love!!!
created my first vibe coded App with this model too
app link : https://ai.studio/apps/drive/1FxLWERq9tNgjjaKvMTaI7cncwibYchUo
generations are Crazyyy
Starting from a single real banana photo to this. A mind blowing model.
The image editing feature is really impressive. I’ve already been experimenting with it for photo compositing, such as generating a possible photo of a couple’s future child based on their pictures.
Why the name? Immediately makes me think "micro-penis."