Google claims to have the most powerful AI version of Photoshop, which is now available to everyone, and the results are truly impressive.

The wildly popular and mysterious image editing model nano-banana has finally shed its "banana peel"!

Just today, Google officially claimed it and stated that this model is actually Gemini 2.5 Flash Image.

The latest wave of official demos has also arrived:

It also demonstrates excellent image reasoning ability:

Now, this model can be used for free on Gemini and Google AI Studio. Meanwhile, the API is available at a price of $0.039 per image.

How does the model perform specifically? Let's take a look~

Excellent Image Editing Ability

The official has provided many interesting usage examples.

In addition to conventional photo - editing methods, it can also combine up to three images to create new things.

Generate surreal art, combine different photo elements, or seamlessly blend objects, colors, and textures.

Or easily realize design ideas.

It can also generate a coherent plot story comparable to a movie scene using several photos.

Before the official confirmation, netizens were already attracted by various private cases of nano - banana and started crazy discussions.

For example, precisely change facial expressions and weather.

Or give a bald man different hairstyles again.

Netizens said: Photoshop is in jeopardy.

After all, this "banana" can not only edit flat images but also perform 2D to 3D conversion. Users only need to provide an ordinary 2D image, and the system can automatically generate the corresponding 3D perspective, with every angle being just right and harmonious.

The most amazing thing is that it can even perform image reasoning. The Pythagorean theorem is a piece of cake, calculating angles is a breeze, and it can instantly recognize the structure of a jellyfish in an image. It's like the AI version of "You Draw, I Guess".

To be honest, what impressed me the most is that when we asked this "banana" to provide a solution for a person to get from point A to point B, it actually generated a ladder!

Well, well, well, the image editor can really think.

We also tested this model ourselves and found it really interesting:

Give it a cat, and the cat can travel around the world.

Prompt: Generate travel photos with this cat as the protagonist.

Prompt: Let the cat take a selfie at the Forbidden City.

If you don't want a real - life scenario, it's just a matter of a single prompt to make Lin Daiyu and Sun Wukong become Marvel heroes and fight side by side~

In addition, its understanding of light and shadow details is quite realistic.

Prompt: A banana is placed on the kitchen counter, with a water cup, a knife, and a cutting board beside it. The sun shines in through the window, and the banana casts a natural shadow.

You can see that the generated result is quite good. The shadow refracted by the sunlight through the water cup and the shadow of the nearby plant are quite reasonable.

It can also well capture the characteristics of a character and change the character's main body without any surprises:

What if Kudo Shinichi didn't drink the shrinking potion that year, but...

The Mysterious Banana Model

Nano Banana first appeared on a website called LMArena, a place where different AI models compete anonymously in a "battle mode".

The reason why Nano - banana became so popular is simple: its performance is really excellent.

This is another state - of - the - art image editing model - and there was no official documentation at the beginning.

However, before the official announcement, some people had already guessed that it was a member of the Google family.

On the one hand, the performance of this model is similar to the recently released multimodal version of Gemini. It's hard to believe that such excellent performance could come from a garage team.

On the other hand, its release strategy - anonymous release, no brand logo, and letting the user community explore on its own - is exactly the same as the way DeepMind early evaluated large language models on anonymous testing platforms.

There is also an additional reason: some developers associated with Google posted strange banana - related content on social media 🍌

Some netizens asked why Google chose to release it secretly, and the official gave this explanation:

They are preparing for a global - scale release.

Netizens: When Will the Complete Version Be Released?

Since the official said so, netizens couldn't sit still.

They all said: Can we use the full version now?

However, although this model performs really well in image generation, it doesn't mean it has no problems -

Especially under the service policies of large companies, some rather frustrating situations can easily occur: