Last night, I saw a post from Master Guizang on WeChat, announcing that the image model on Google’s AI Studio has been updated to Gemini 2.5 Flash Image Preview, which is said to be Nano Banana.

I quickly opened my computer and tried it out. After testing the first image, I was convinced that this is indeed Nano Banana.

So, I searched for news from Google, and it was indeed released.

From the information provided, it was mentioned that “in the early preview, people praised it highly, stating it is the most popular image editing model in the world,” which confirms that it is indeed the previously mysterious Nano Banana model. The reason I am so focused on it is that it is incredibly powerful.

Currently, it ranks first in the Image Edit category on lmarena, with a score that is far ahead.

This is an article I just wrote a couple of days ago on https://lmarena.ai/ evaluating Nano Banana.

Nano Banana is the image editing model with the strongest character consistency.

I didn’t test enough last time, so I want to test it again today.

Where to Use It

The official provides several ways:

• Gemini App
• Google AI Studio
• Vertex AI
• Google Open API Interface

Unofficial access points: (only listing a few I know)

• Flowith
• Adobe Firefly
• ComfyUI

Pricing

Gemini 2.5 Flash Image Preview is a raw text and image model.

In image generation mode:

• Input: $0.3 / 1M Tokens
• Output: $30 / 1M Tokens

One image costs about 30 cents, and currently, using the official website/app interface is free, with charges only for API calls.

Currently, the first month of Gemini is free, so hurry up and take advantage of it!!!

With the pricing and usage methods sorted out, how powerful is it? Let’s let the cases speak for themselves.

One Shot

Next, I will only use two original images for all test cases to explore how strong the editing capabilities are. Continuing from my previous test case, this time I will use Iron Man along with an image of Captain America.

The original images are as follows:

Without further ado, let’s directly look at the prompts and the corresponding images (all images below are based on the original images or newly generated images as reference).

Epic movie still: A massive alien warship tears through the clouds, threateningly appearing over the city, its surface covered with complex geometric structures and purple light, casting shadows. Iron Man from image 1 and Captain America from image 2 stand on the ruins, facing away from the camera, looking at the alien warship, in a style consistent with the previous scene, 16:9 widescreen.

Epic movie still, close-up, Iron Man and Captain America look at each other, Iron Man removes his mask, Captain America still wears his mask, both showing determined expressions, with a massive alien warship in the background, 16:9 widescreen.

Epic movie still, medium shot, Iron Man’s feet and hands simultaneously emit dazzling blue and white lasers, the lasers cone downwards, releasing a strong energy beam. A 45-degree upward shot from the lower right corner captures Iron Man’s back, charging towards the alien warship in the sky, emphasizing speed and power, in a style consistent with the previous scene. 16:9 widescreen.

Epic movie still, during Iron Man’s flight towards the warship, dozens of insect-like mechanical soldiers fly out from the alien warship into the sky.

Epic movie still, high-altitude combat close-up, Iron Man skillfully dodges the energy fire from the insect-like mechanical soldiers and retaliates fiercely with palm cannons and missiles. 16:9 landscape.

Epic movie still: Captain America accurately deflects the alien purple energy beam with his shield. The style is consistent with the previous confrontation scene, 16:9 widescreen.

Movie still: Under Iron Man’s fierce attack, the alien warship’s side armor ruptures, exposing its structure, with sparks and smoke emerging, creating a breakthrough for the attack. The style is consistent with the previous scene, 16:9 widescreen.

Epic movie still, Iron Man is on the left side of the frame, the alien warship is in the upper right side of the frame, the circular reactor on Iron Man’s chest emits a powerful shock beam, striking the alien warship in the upper right corner from the left side of the frame. The beam is dazzling and full of energy, 16:9 widescreen.

A close-up of the alien warship in the upper right corner, the beam hits the warship, and the left side of the warship explodes.

Epic movie still, showcasing Iron Man and Captain America back-to-back, fiercely resisting the alien insect-like mechanical soldiers. The scene is grand, the special effects are stunning, and the sense of action is strong, continuing the visual style of the movie.

Epic movie still, the alien warship in the sky explodes, and all the alien insect-like mechanical soldiers also explode simultaneously. The scene is grand, the special effects are stunning, and the sense of action is strong, continuing the visual style of the movie.

With just two images, so much imagination can be generated. The key point is that from these more than ten images, we can see how powerful the Gemini 2.5 Flash Image model is. The reference images provided can almost all be restored, allowing one person to create blockbuster material.

However, it may be because there are more training materials for Iron Man and Captain America, allowing the model to restore them better.

Of course, it is not without flaws. Since version 2.0, there has been a problem where the generated images are not high definition, but it is much better than 2.0.

I want to test another case, to test multi-image fusion again!

Multi-Image Fusion

Let’s directly look at the output effect.

Color the shoes in image 1 with the style of image 2, the shoes need to have elements of blue sky, white clouds, mountains, oceans, and boats, with a pure white background.

A man wears the shoes from image 1.

The man in image 1 takes off his coat and changes into the clothes from image 2.

The man in image 1 changes into the pants from image 2.

The man puts on the hat from image 2.

High-definition close-up of the man’s face.

Stitch from Lilo & Stitch becomes a 3D model, standing on the man’s left shoulder.

The camera pulls back to give a wide shot, on a fashion runway.

Whether it is style transfer, image fusion, outfit change, high-definition restoration, composite images, or 3D effect conversion, it has all been done very well. This model has completely convinced me that the future is a model-as-a-service future. Why use an app for something that can be done with just a prompt?

END

To summarize, over the past two days, I have felt that Nano Banana has the following capabilities:

• Blockbuster generation
• Character consistency
• Local editing
• Multi-character interaction
• Style transfer
• Outfit change and fusion
• High-definition restoration
• 3D conversion

There are still some scenarios I haven’t tested, and I will continue to share more usage scenarios in the future if I have the opportunity.

As I tested, my biggest feeling is that Gemini 2.5 Flash Image has truly lowered the threshold for image editing to an extremely low level. In the past, you might have needed Photoshop, plugins, and repeated adjustments of layer parameters, but now, with just a prompt or even a reference image, you can generate images comparable to blockbuster films. More importantly, it not only stops at generating new images but also integrates editing, fusion, outfit changes, restoration, and 3D conversion into a new creative paradigm.

What does this mean? It means that the future of content production will be completely changed: apps will no longer be the only entry point; model-as-a-service will be the endpoint. Designers, video creators, brand owners, and even ordinary users can directly call models to complete their creativity without having to go through a complex toolchain. Creativity itself becomes the core, and execution is handed over to the model.

Of course, the model still has imperfections, such as resolution and Chinese text rendering, but this does not prevent it from redefining industry standards. Looking back today, the mysterious model known as Nano Banana has finally unveiled its veil, and we may be on the brink of a new era of visual creation.

So, if you haven’t tried it yourself yet, hurry up and get started. You will find that the future is closer than you think.

Xia Wei | Your dedicated AI pocket

Looking forward to the future, we can use AI to liberate productivity together and become super individuals.