2+ on paid plans
Watermark
Removed on paid plans
Display public

Nano Banana 2 生成结果

Nano Banana 2 AI 图像生成与编辑器

Google 最新 Nano Banana 模型,发布于 2026 年 10 月 6 日:文生图与图生图融合于一体,支持最多 14 张参考图,1K、2K 或 4K 输出,每张仅需 12 积分起。

什么是 Nano Banana 2.1?

Nano Banana 2.1 is Google's newest image generation and editing model, announced on October 6, 2026 with the Gemini API model ID gemini-nano-banana-2.1. Google describes it as an update to Nano Banana 2 (Gemini 3.1 Flash Image) that keeps Flash-level speed and cost while improving visual quality, prompt adherence, character consistency and text rendering. Its model card says it is based on Gemini 3.6 Flash, and Google positions it as the more efficient counterpart to Nano Banana Pro.

According to Google's Gemini API model page, the main changes are better realism at 1K, 2K and 4K; fixed tiling artifacts on extreme ratios (1:4, 4:1, 1:8, 8:1) at 2K and 4K; more accurate text and infographic layouts; and up to 14 reference images, with consistency for up to 4 characters and 10 objects.

On Flux1 AI, Nano Banana 2.1 has been live since October 9, 2026 and is the default model in the generator above. You write a prompt, optionally add up to 14 reference images, pick a ratio and a resolution, and get one image in about 60–90 seconds. The same model is available to developers on the Flux1 AI API at the same credit prices.

开发者
Google DeepMind
发布时间
October 6, 2026
支持模式
Text-to-image and image editing
参考图像
每次最多 14 张, no extra cost
Flux1 AI 输出规格
1K, 2K or 4K, one image per run
画面比例
Auto, 1:1 to 21:9, plus 1:4, 4:1, 1:8, 8:1
典型耗时
About 60–90 s
点数资费
12 (1K), 16 (2K) or 21 (4K) credits

What Nano Banana 2.1 does well

Google's four headline upgrades, and how each one works on this page.

Cleaner panoramas and tall strips

Google reports better realism at all three resolutions and fixed tiling artifacts on very wide or tall images at 2K and 4K. Here you can pick 1:4, 4:1, 1:8 or 8:1 directly: a 1:8 run at 1K came back at 352 × 2928 px, a 16:9 run at 2K at 2752 × 1536 px.

Text and infographics

Google lists enhanced text rendering and infographic layout accuracy as a key update, and its model card puts 2.1 ahead of Nano Banana 2 and Pro on its infographic tests. Short headlines in quotes, numbered lists and labeled diagrams are where it pays off; render at 2K or 4K when small text matters.

Up to 14 reference images

Upload product shots, faces, logos or style frames and tell the model which image supplies what. Google tunes 2.1 to keep up to four characters and ten objects consistent. References cost nothing extra here, so a 14-image composite costs the same as a plain prompt at the same resolution.

Generate and edit in one model

There is no separate edit mode. With no reference, 2.1 creates from text; with references, it edits or combines them. Set the ratio to "auto" with a reference and the output keeps the input's proportions, which is what you want for background swaps and retouching.

如何使用 Nano Banana 2.1

Four steps from prompt to download. Credits are only spent when a run succeeds.

  1. 1

    Write the prompt

    Nano Banana 2.1 is already selected in the panel above. Describe subject, action, setting, lighting and framing in plain sentences, and put any text that must appear in quotes with its position.

  2. 2

    Add references (optional)

    Upload up to 14 images to edit or combine. Refer to them by number in the prompt, for example "the jacket from image 2 on the person from image 1".

  3. 3

    Choose ratio and resolution

    Pick a ratio from 1:8 to 8:1, or "auto" to follow your first reference. Choose 1K for drafts, 2K for most web use, 4K for print or crops. The credit cost updates before you press Generate.

  4. 4

    Wait, then iterate

    A run usually takes 60–90 seconds. You can leave the page; the image lands in your history. Reuse a result as a reference to change one detail, or rerun your best 1K draft at 4K.

Prompt tips for Nano Banana 2.1

Since each run takes about a minute, a specific prompt saves more time here than on faster models. Name the scene, quote the text, number the references.

Vague prompt
Prompt that works
A cool banner for my shop
8:1 web banner for a ceramics shop: three mugs on a pale oak shelf, soft morning light from the left, headline "Handmade in Lisbon" on the right third
Put me in Paris
Keep the face, hair and glasses from image 1. Place her on a café terrace in Paris at dusk, seated, holding an espresso, 50mm lens look
An infographic about sleep
Vertical 1:4 infographic titled "Better Sleep": 5 numbered tips with simple icons, navy and cream palette, one short line of text per tip
Make it better
Change only the background to a white studio sweep. Keep the product, label text and camera angle exactly as in image 1

Nano Banana 2.1 vs other models on Flux1 AI

All of these run on the same credit balance, and the first four are in this generator's model menu.

模型适用场景Max outputReferences典型耗时Credits / image
Nano Banana 2.1Infographics, extreme ratios, multi-reference composites4KUp to 1460–90 s12–21
Nano Banana 2Fast everyday work up to 4K4KUp to 1410–20 s10–21
Nano Banana ProFace likeness, in-scene text, final assets4KUp to 1410–15 s14–28
Nano Banana 2 LiteQuick 1K drafts and social graphics1KUp to 105–10 s6
GPT Image 2.5A second opinion from OpenAI's model4KUp to 1430–60 s12–24

Nano Banana 2.1 vs Nano Banana 2

Google calls 2.1 an update to Nano Banana 2: same 1K/2K/4K range and 14-reference limit, with better realism, prompt adherence, consistency and text, and cleaner extreme ratios. On Flux1 AI, 2.1 costs 12 / 16 / 21 credits against 10 / 14 / 21 for Nano Banana 2, and it is slower here: about 60–90 seconds versus 10–20. Use Nano Banana 2 for fast iteration, 2.1 when text, layout or a wide panorama has to be right.

Nano Banana 2.1 vs Nano Banana Pro

On Google's human-preference tests in the 2.1 model card, 2.1 scored above Pro in every listed category, and Google calls it the more efficient counterpart to Pro. Early independent same-prompt tests are more mixed and still favor Pro for matching a reference face and for signs inside a scene. Nano Banana Pro costs 14 credits at 1K or 2K and 28 at 4K, and returns in 10–15 秒 here.

Nano Banana 2.1 pricing

You see the credit total before you press Generate, and failed runs are refunded automatically.

RunWhat you getCredits
1KText-to-image or an edit, about 1 megapixel, good for drafts and social posts12
2KTwice the width and height, for web pages, ads and readable small text16
4KUp to 4096 × 4096 px at 1:1, for print, packaging and large displays21
ReferencesUp to 14 images per run at any size, no surchargeIncluded

How far a monthly plan goes on Nano Banana 2.1

  • Basic · 300 credits/mo

    ≈ 25 images at 1K, or 18 at 2K

  • Pro · 800 credits/mo

    ≈ 66 images at 1K, or 50 at 2K

  • Max · 2000 credits/mo

    ≈ 166 images at 1K, or 125 at 2K

Credits are shared across every image and video model, so 2.1 finals, quick drafts on Nano Banana 2 Lite and a few videos all come out of one balance.

Good to know before you start

What this page does differently from Google's own apps, and where the model still slips.

Slower here than Nano Banana 2

A run usually takes 60–90 seconds on Flux1 AI, against 10–20 seconds for Nano Banana 2. Draft ideas on a faster model, then use 2.1 for the version you keep.

No Thinking or search controls

Google's API offers Thinking levels (minimal, medium, high) and grounding with Google Web and Image Search. Neither is exposed here, so the model does not look up current facts for you: write verified numbers, names and dates into the prompt.

No mask, brush or chat editing

Google highlights mask-based and multi-turn conversational editing. This generator has no brush or selection tool and no chat thread: you edit with references and one instruction per run, and to continue, reuse the result as a reference.

One image per run

Each run returns a single image. For variations, run the same prompt again or change one phrase at a time so you can see what made the difference.

Proofread the text

Google's model card for 2.1 lists blurry small text at 1K, long paragraphs and left/right confusion as known weak spots. Keep on-image text short, render at 2K or 4K when it is small, and check every word before publishing.

Content rules apply

Explicit or "spicy" sexual content is not supported. Google's safety filters can block prompts or references that break the content policy, and a blocked run is refunded.

Nano Banana 2.1 常见问题解答

Can I use Nano Banana 2.1 on Flux1 AI?

Yes. Nano Banana 2.1 has been live on Flux1 AI since October 9, 2026 and is the default model in the generator on this page. Write a prompt, add references if you have them, pick 1K, 2K or 4K and press Generate. The model menu also offers Nano Banana 2, Nano Banana 2 Lite and Nano Banana Pro.

Is Nano Banana 2.1 free?

It runs on credits: 12 at 1K, 16 at 2K and 21 at 4K, with reference images included. Credits come with any plan and are shared across all models. Failed or blocked runs are refunded automatically, so you only pay for images you get.

When was Nano Banana 2.1 released?

Google announced Nano Banana 2.1 on October 6, 2026 and began rolling it out across the Gemini app, Google AI Studio, the Gemini API and the Gemini Enterprise Agent Platform, among other products. It reached Flux1 AI on October 9, 2026.

什么是 the difference between Nano Banana 2.1 and Nano Banana 2?

2.1 is Google's update to Nano Banana 2 (Gemini 3.1 Flash Image) and is based on Gemini 3.6 Flash. Google reports better realism, prompt adherence, character consistency and text, plus cleaner 1:8 and 8:1 images at 2K and 4K. Output sizes and the 14-reference limit are the same. On Flux1 AI, 2.1 takes about 60–90 seconds per image, Nano Banana 2 about 10–20.

Is Nano Banana 2.1 better than Nano Banana Pro?

On Google's own preference tests, 2.1 scored above Pro in every listed category, and it costs less at 4K here (21 versus 28 credits). Independent tests still favor Pro for matching a reference face and for signs inside a scene, and Pro is faster on this site. Try the same prompt on both; both are in the model menu above.

Is there a Nano Banana 2.1 API?

Yes. On the Flux1 AI API, use model nano-banana-2-1 for text-to-image and edits with up to 14 references, at the same credit prices as this page; the API docs cover requests and results. Google also offers it directly as gemini-nano-banana-2.1 in the Gemini API.

What happens to Nano Banana 2?

Google has deprecated gemini-3.1-flash-image in the Gemini API and names Nano Banana 2.1 as its replacement. As of October 9, 2026, Google's deprecations page shows no shutdown date. Nano Banana 2 stays available on Flux1 AI for now, at 10–21 credits.

Can I download Nano Banana 2.1 images?

Yes. Download any result from its card. Each image is also saved to your history, so you can reuse it as a reference later. Files come at the resolution you chose: 1K, 2K or 4K.

Create with Nano Banana 2.1

Write a prompt, add up to 14 reference images, and get a 1K, 2K or 4K image in about a minute.