Google Nano Banana (Gemini 2.5 Flash Image)
Google Nano Banana 是一个 AI 图像模型的代号，该模型正式名称为 Gemini 2.5 Flash Image，由 Google DeepMind 开发。它已集成到 Google 的 Gemini 应用中，并提供高级图像生成和编辑功能，尤其侧重于保持角色和场景的一致性。在下方的 HIX AI 上试用 Google Nano Banana AI 图像模型！
Google Nano Banana AI 图像模型的关键特性
- 角色和风格一致性极佳：保持原始图像中角色、物体或风格的一致性
- 多图像融合：将不同的输入图像融合为一张新的无缝图像
- 视觉推理与世界知识能够处理涉及推理的复杂图像生成。
角色和风格一致性极佳
保持角色和风格的一致性是Nano Banana （Gemini 2.5 Flash Image）最显著的特点之一。Nano Nano Banana AI 图像模型在这方面表现出色，它使用户能够在不同的场景或视角中放置相同的角色或产品，而不会丢失身份或视觉风格。
以下是一些示例，说明其特征和风格的一致性：
制作动漫人物的动作人偶
Nano Banana AI图像模型可让您根据动漫人物的图像创建逼真的动作人偶照片。以下是一些示例。
| 输入图像和提示 | 输出图像 |
 提示： 创作一个 1/7 比例的写实风格动作人偶，人物形象与图片一致，置于真实环境中。人偶放置在电脑桌上。电脑屏幕上显示的是人偶的 C4D 建模过程。电脑屏幕旁边是一个万代风格的塑料玩具包装盒，上面印有原始插图。电脑桌上还有制作人偶的工具，如画笔、颜料、小刀等。 |
 |
非卖品
— TianZ (@Tian416) 2025年8月23日
*由Nano Banana生成pic.twitter.com/b1OeLF9xOI
训练纳米香蕉模型的人真是太棒了。pic.twitter.com/m9SF6oBLmm
— Tepobeso (@beet_oops) 2025年8月17日
添加或编辑特定元素
Nano Banan 能够更强大地帮助您修改图像中的特定元素，而不会影响图像的周围区域、样式、透视和光照。
| 输入图像和提示 | 输出 |
 提示：让图片中的猫戴上牛仔帽，手里拿着一根香蕉。 |
 |
 提示：将连衣裙的颜色改为蓝色。 |
 |
多图像融合
Nano Banana AI图像模型的另一个突出功能是能够将不同的输入图像融合为一张新的无缝图像。这对于组合各种元素非常有用，例如将产品放置到新环境中、重新设计环境或创建融合多个参考的广告。融合尊重视觉连贯性和光照，使生成的图像看起来自然逼真。
| 输入图像 |
 |
 |
 |
| 提示 | 将笔记本电脑放在办公室的桌子上。让女子坐在照片左侧的椅子上，使用笔记本电脑工作。 | ||
| 输出 |
 |
||
视觉推理与世界知识
Nano Banana受益于 Gemini 的深厚世界知识，能够处理超越简单美学的复杂图像编辑和生成任务。
它可以理解手绘图表，遵循单个提示中的多步指令，并以语义感知的方式解释场景。这使其在教育工具、交互式应用程序以及需要上下文和推理的实际问题解决场景中更具通用性。
| 输入图像和提示 | 输出 |
 提示：求解 x，并将 x 替换为蓝色字体书写的正确值。 |
 |
Nano Banana 1对比Nano Banana 2
如需了解更多详情，请参阅我们对初代Nano Banana和其后续产品Nano Banana 2 的对比。
| 方面 | Nano Banana 1 | Nano Banana 2 |
| 生成速度 | ★★★★ | ★★★★★ |
| 整体美感 | 有能力 | 精致 |
| 编辑功能 | 支持基本的文本转图像和标准图像编辑功能。 | 在聊天中原生引入最先进的构图和高级风格转换功能。 |
| 提示遵守 | ★★★ | ★★★★ |
| 文本渲染 | 经常难以处理复杂的排版和较长的短语。 | 在将清晰、准确的文本无缝集成到图像中方面，可靠性显著提高。 |
| 超写实主义 | 好 | 令人信服 |
| 生态系统角色 | 作为可靠的基线生成图像模型。 | 为Gemini生态系统中的快速、深度集成的图像生成和复杂编辑功能提供支持。 |
如何在HIX AI上使用Google Nano Banana
- 1
选择Nano Banana模型
前往HIX AI图生图 AI 生成器，然后选择Google Nano Banana模型。
- 2
输入提示词
上传图片并输入文本指令。
- 3
开始生成
点击“生成”，即可快速得到生成图像。
关于 Google Nano Banana (Gemini 2.5 Flash Image) 的 YouTube 视频
关于 Google Nano Banana (Gemini 2.5 Flash Image) 的 Reddit 讨论
Google really raised the bar with nano banana, scary how good and accurate it is.
byu/Mephistophilis44 insingularity
Reddit 上 localllama
社区的帖子
Reddit 上 singularity
社区的帖子
关于 Google Nano Banana (Gemini 2.5 Flash Image) 的 X 帖子
🚨🍌Big Reveal: who was "Nano Banana?"
— lmarena.ai (@lmarena_ai) August 26, 2025
The anonymous model, “nano-banana,” that caught the world's attention with its ability to follow complex instructions, preserve character identity, and maintain contextual details was:
Gemini-2.5-Flash-Image-Preview by @GoogleDeepMind 🍌✨… https://t.co/O1wpnbxaK3 pic.twitter.com/U4va9hJabk
AI just went B-A-N-A-N-A-S…
— Amira Zairi (@azed_ai) August 27, 2025
Meet Nano Banana, a tiny but insanely powerful AI model that upgrades your workflow like a cheat code 🍌
It’s fast, lightweight, and feels illegal to be free.
Here’s how to use it for FREE, with 13 use cases + prompts.
Bookmark this for later pic.twitter.com/fGRY9yX8LB
nano-banana attempt to restore the first photo ever taken pic.twitter.com/k2QiwkYGhI
— Denis Shiryaev 💙💛 (@literallydenis) August 26, 2025
Nano-banana is probably the best image editing model right now
— @levelsio (@levelsio) August 26, 2025
- Left: my input photo
- Prompt: "show the man holding this in the bed"
- Right: nano-banana's output
Sergey Brin + @OfficialLoganK are really saving Google! https://t.co/4JIZ95tlFq pic.twitter.com/S30IWUsYyv
I found a method with Nano Banana that is more accurate than uploading individual photos. Create a collage with everything on a single image and label each item/person. Then when you type a prompt mention them.
— Travis Davids (@MrDavids1) August 27, 2025
Here is a direct comparison that I could not get to work accurately… pic.twitter.com/yIYHqQjVEm
Introducing Gemini 2.5 Flash Image (aka nano-banana), our SOTA image generation and editing model 🍌
— Logan Kilpatrick (@OfficialLoganK) August 26, 2025
As you might have already seen, this model excels at character consistency, creative edits, and has Gemini's world knowledge! pic.twitter.com/kJQVJcMPE6
+1 on this
— Linoy Tsaban🎗️ (@linoy_tsaban) August 27, 2025
Nano banana's editing is really next level, but for T2I it's not as mind blowing
for this one for example I actually like Qwen better (left/top) than gemini (right/bottom) https://t.co/WFHhKBfyRn pic.twitter.com/GLOFSntamn
Having some fun with Nano Banana from Google. I can finally create my adult animated series, "The Life of an AI Bro." 🍌🍌🍌 pic.twitter.com/XFrX0R4OEf
— Dave Clark (@Diesol) August 26, 2025
I’ve been testing Nano Banana in Google Gemini, and yeah — no doubt it’s the best image editor out there.
— OscarAI (@Artedeingenio) August 26, 2025
What I find most interesting is how good it is for changing a character’s expression and gestures.
Here I used the image of Jinx you see on the left, and I tweaked her… pic.twitter.com/xkkJpXwnJ0
Nano-banana is out! 🤏🍌
— Justine Moore (@venturetwins) August 26, 2025
I'm really impressed by this model's language understanding - it can do more with less guidance.
You can say things like "continue this story" and it will write a narrative and then create the image(s) for you.
(I'm running it on @krea_ai) pic.twitter.com/qNxdY816c3
1/ INSANE (LONG) THREAD.
— Rodrigo Bressane (@bressane) August 26, 2025
Nano Banana 🍌 for photo restoration and colorization.
Nothing has ever been like this before. I spent countless hours working on photo restoration and colorization solutions long before AI was a thing. Nothing compares to this. Truly remarkable.… pic.twitter.com/HBFN4WwTw1
Playing around more with nano-banana, trying to create a thumbnail based on some examples. When the ask is ambiguous, as with all AI you get a sloppy/incoherent response. 1/ pic.twitter.com/aWaEaEGhub
— Machine Learning Street Talk (@MLStreetTalk) August 27, 2025
nano banana finally makes it easy to do what normal people want to do with images (just say what you want and have it happen with high fidelity)
— David Gasca (@gasca) August 26, 2025
this image explaining how the prompt works was also made by the model (you can see it wasn't perfect though...) pic.twitter.com/dYFZ10g7sm
Nano Banana is kinda crazy, Google did a solid job with this one!
— Ege (@egeberkina) August 26, 2025
I’ll be sharing more of what I’ve created with Nano Banana soon 🍌 pic.twitter.com/ls2xHMTh3q
The results from Google’s new image model “nano-banana” are truly amazing.
— CHOI (@arrakis_ai) August 14, 2025
I simply asked it to create a photo of someone looking straight ahead.@OfficialLoganK pic.twitter.com/LWH1jM1Pki
It's official! Nano-Banana was Google all along (as if it was a big secret - lol). I did have early access but wasn't allowed to officially confirm it was Google until today.
— Matt Wolfe (@mreflow) August 26, 2025
But the really good news is that it's available to everyone now in the Gemini App! No more LMArena… pic.twitter.com/uxoZBOYF9D
常见问题解答
