Nano Banana 2.1 — Google's New Workhorse Image Model, Online

Powered by gemini-nano-banana-2.1 on a Gemini 3.6 Flash base. Sharper text and infographics, 8:1 panoramas without tiling, up to 14 reference images, 4K output — at half the price of Nano Banana 2.

Image Generator
0 / 5000

Advanced Options

Cost 3 credits

New accounts get 3 free credits — exactly enough for your first generation.

Image Preview
Infographics that get the numbers right
What is Nano Banana 2.1?

Nano Banana 2.1 (model ID gemini-nano-banana-2.1) is Google's image generation and conversational editing model released on 6 October 2026. It is built on Gemini 3.6 Flash and replaces Nano Banana 2 as the high-efficiency workhorse of the family: Google now recommends it for all new projects. On Google's own model card it beats the previous Nano Banana 2 on every reported benchmark and edges past Nano Banana Pro on several — overall preference, infographic design, multi-character consistency and mask-based editing — while staying on the Flash side of the speed and cost curve. imagesv2.ai runs it in the browser: the generator above is already set to it, no Google account or API key required.

Four things Nano Banana 2.1 does better than Nano Banana 2

Nano Banana 2.1 AI generated infographic comparing the tallest mountain on each continent with correct elevations

Infographics that get the numbers right

Infographic factuality is the single biggest jump on the model card: 0.521 for Nano Banana 2.1 against 0.179 for Nano Banana 2 and 0.265 for Pro. Give it real figures and a chart type and it lays them out legibly, keeps labels attached to the right bars, and stops inventing decimals. Still proofread before you publish — but you are now checking, not redrawing.

Prompt
A clean editorial data infographic titled exactly "TALLEST MOUNTAINS ON EACH CONTINENT", landscape layout. Seven stylised mountain silhouettes drawn side by side, scaled proportionally by real height, each labelled with its correct name and elevation in metres: Everest 8,849 m, Aconcagua 6,961 m, Denali 6,190 m, Kilimanjaro 5,895 m, Elbrus 5,642 m, Vinson 4,892 m, Puncak Jaya 4,884 m. A small continent label under each mountain. Muted navy, sand and terracotta palette on an off-white paper background, thin grid lines, modern sans-serif type, generous margins. All numbers and names spelled exactly as given and perfectly legible.
Nano Banana 2.1 AI generated 8:1 panoramic photograph of a Nordic fjord at golden hour with a continuous mountain ridge

8:1 panoramas without the tiling

Nano Banana 2.1 adds the extreme ratios — 4:1, 1:4, 8:1 and 1:8 — and fixes the repeated-pattern artefacts that wide canvases used to show. A ridge line, a skyline or a product shelf now runs continuously across the whole strip. Useful for website hero banners, trade-show backdrops, ultrawide wallpapers and social covers that used to need two generations and a seam.

Prompt
An ultra-wide 8:1 panoramic landscape photograph of a Nordic fjord at golden hour: a continuous mountain ridge running the full width of the frame, snow on the peaks, a calm fjord mirroring the sky below, a tiny red fishing village on the shore near the right third. The ridge line and the water flow continuously across the whole strip with no repeated or tiled sections and no seams. Warm low sunlight from the left, long shadows, crisp detail, natural colour.
Nano Banana 2.1 AI generated two-panel comic keeping the same four characters consistent across both panels

A whole cast, not just one character

Multi-character consistency scores 1106 Elo on the model card, a hundred points above Nano Banana Pro. Four distinct people stay themselves across panels — same faces, same hair, same clothes — which is what a storyboard, a comic, a children's book or a brand mascot series actually needs. Add reference photos in image-to-image mode and it holds up to 14 of them at once.

Prompt
Two side-by-side comic panels with a thin white gutter, semi-realistic illustrated style with a warm cinematic colour grade. The SAME four characters appear in both panels and stay recognisably identical in face, hair and clothing: (1) a tall man with a grey beard and a navy fisherman sweater, (2) a young woman with a black bob haircut and a mustard yellow raincoat, (3) a teenage boy with curly red hair, freckles and a green hoodie, (4) an elderly woman with white hair in a bun, round glasses and a plum cardigan. LEFT panel: the four of them huddle around a paper map on a wooden harbour pier at dawn. RIGHT panel: the same four stand on the deck of a small boat at midday, the bearded man steering, the boy pointing at the horizon.
Nano Banana 2.1 AI generated bilingual English and Chinese café menu board with sharp small lettering and prices

Small bilingual text that survives a zoom

Text rendering was already a Nano Banana 2 strength; 2.1 pushes it into the small sizes — menu lines, price tags, footers, mixed Latin and CJK on the same board. The characters come out correctly formed and evenly spaced, so a bilingual menu, packaging mock-up or signage concept can go straight into a deck. Dense paragraphs are still the hardest case for any image model.

Prompt
A straight-on photograph of a bilingual café menu board hanging on a warm plaster wall. Cream board with dark green hand-painted lettering. A large heading reading exactly "MORNING MENU 早餐菜单". Below it, four menu lines, each in English then Chinese with a price on the right: "Soy Milk 豆浆 ¥6", "Fried Dough Stick 油条 ¥4", "Steamed Bun 包子 ¥5", "Congee 白粥 ¥7". At the bottom a small line reading exactly "Open daily 7:00 – 11:00 每日营业". Every Latin letter and every Chinese character correctly formed, evenly spaced and sharp, including the small bottom line.
Key Features
  • Model ID gemini-nano-banana-2.1, built on Gemini 3.6 Flash — released 6 October 2026
  • 1K, 2K and 4K output; 2K returns about 2750 px and 4K about 4100 px on the long edge
  • Full aspect-ratio range plus the new extremes: 4:1, 1:4, 8:1 and 1:8, with wide-canvas tiling fixed
  • Up to 14 reference images in one task, including up to 10 high-fidelity object references
  • Conversational multi-turn editing with mask / ink region edits (1049 Elo vs 965 for Nano Banana 2)
  • Thinking levels minimal / medium / high — reasoning stays on, so layouts and text are planned before drawing
  • Google Web Search and Image Search grounding available through the Gemini API for real-world subjects
  • Every image carries Google's invisible SynthID watermark; Nano Banana 2 is now marked for migration to 2.1
Pricing and Credits

Google priced Nano Banana 2.1 at about half of Nano Banana 2: $0.0336 per 1K image, $0.0504 at 2K and $0.113 at 4K. On imagesv2.ai that is 40, 60 and 120 credits respectively, plus a 20-credit surcharge when you upload an image to edit. The cost is shown before every generation, and new accounts get free credits on sign up.

What Nano Banana 2.1 is great at

Infographics and data visuals

Charts, comparison graphics and explainer slides from real numbers. The factuality jump means the labels and values you give it are the ones that end up on the canvas.

Product and e-commerce imagery

Keep the product exactly as photographed while the scene, lighting and backdrop change — and run the whole catalogue at Flash speed and the new lower price.

Storyboards, comics and character series

A fixed cast across many frames. 2.1's multi-character consistency makes sequential content practical rather than a lottery.

Banners, panoramas and ultrawide covers

Hero images, event backdrops and 8:1 social headers in one pass — no seams, no stitching two generations together.

Signage, menus and packaging mock-ups

Bilingual boards, labels and price lists where the small type has to read correctly in both Latin and CJK scripts.

Fast iteration and in-place edits

Say what to change and keep the rest. Mask-style region edits and multi-turn conversation let you fix one corner instead of regenerating the whole image.

How to use Nano Banana 2.1
  1. 1

    Open the generator

    The generator above already has Nano Banana 2.1 selected. Sign in to claim your free credits — no card, no API key.

  2. 2

    Describe the job, not just the subject

    Say where the image will go — banner, chart, menu board, product listing. Name the exact text and numbers you want and put them in quotes; 2.1 is good at reproducing what you spell out.

  3. 3

    Pick ratio and resolution

    Choose the aspect ratio (including 8:1 and 1:8) and 1K, 2K or 4K. Settle the composition at 1K, then re-run only the keeper at 4K — the credit cost updates as you switch.

  4. 4

    Edit instead of restarting

    Close but not right? Switch to image-to-image, upload the result or your references, and ask for the single change you need — "same four people, move them indoors", "keep the layout, prices in euros".

Nano Banana 2.1 vs Nano Banana 2
Nano Banana 2.1Nano Banana 2
Model IDgemini-nano-banana-2.1 (Gemini 3.6 Flash)gemini-3.1-flash-image (Gemini 3.1 Flash)
StatusCurrent — recommended for new projectsMarked for migration
Overall preference (Elo)1050990
Infographic factuality0.5210.179
Multi-character consistency (Elo)1106978
Mask / ink editing (Elo)1049965
Aspect ratiosStandard set + 4:1, 1:4, 8:1, 1:8, tiling fixedStandard set
Resolutions1K / 2K / 4K0.5K / 1K / 2K / 4K
Price per 1K image≈ $0.0336 (40 credits)≈ $0.067 (70 credits)
Nano Banana 2.1 — Frequently Asked Questions
Common questions about Nano Banana 2.1 and imagesv2.ai. Have more questions? Contact us at support@imagesv2.ai.

Nano Banana 2.1 is Google's image generation and editing model with the API ID gemini-nano-banana-2.1, released on 6 October 2026. It is built on Gemini 3.6 Flash and is the direct successor to Nano Banana 2 (Gemini 3.1 Flash Image): same role as the fast, affordable workhorse of the family, with better visual quality, text, infographics, multi-character consistency and editing. You will also see it written nano banana 2.1, nano-banana-2.1 or Nano Banana 2.1 AI — same model. imagesv2.ai runs it in the browser with no Google account or API key.

On Google's model card Nano Banana 2.1 leads Nano Banana 2 on every reported metric: overall preference 1050 vs 990 Elo, infographic design 1048 vs 961, infographic factuality 0.521 vs 0.179, general editing 1026 vs 938, multi-character consistency 1106 vs 978 and mask/ink editing 1049 vs 965. It also adds the 4:1, 1:4, 8:1 and 1:8 aspect ratios, fixes the tiling artefacts on wide canvases, and costs about half as much per image. Google now recommends new projects use 2.1 and has marked Nano Banana 2 for migration.

On several benchmarks, yes — the model card shows 2.1 ahead of Pro on overall preference (1050 vs 935), infographic design, multi-character consistency (1106 vs 1011) and mask editing. Pro (Gemini 3 Pro Image) is still the larger model and remains the safer choice when a single frame has to be as precise as possible, and it costs 140 credits per image here against 40 for 2.1. For most production work 2.1 is now the sensible default; both sit in the model picker above so you can compare on the same prompt.

Free credits land in your account when you sign up on imagesv2.ai — no card, no API key. After that Nano Banana 2.1 costs 40 credits per 1K image, 60 at 2K and 120 at 4K, with a 20-credit surcharge when you upload an image and edit it. That follows Google's own list price of $0.0336, $0.0504 and $0.113 per image — roughly half of Nano Banana 2 at every resolution.

Nano Banana 2.1 is faster and considerably cheaper, and it is now very strong at infographics, multi-character scenes, wide panoramas and bilingual text. GPT Image 2 is OpenAI's flagship: more exact on long, detailed briefs and on dense typographic layouts, at a higher price per image. Produce at volume with 2.1; when one frame has to be perfect, run the same prompt through GPT Image 2 as well. imagesv2.ai supports both from one picker.

1K, 2K and 4K — the 512 px option that Nano Banana 2 had is not available on 2.1. In our tests a 16:9 request at 2K came back 2752×1536 and a 1:1 request at 4K came back 4096×4096. Aspect ratios cover the usual set (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 5:4, 4:5) plus the new extremes 4:1, 1:4, 8:1 and 1:8, all selectable in the generator above.

Google's model accepts up to 14 reference images in one task, including up to 10 high-fidelity object references, and holds several characters consistently across them. The uploader on imagesv2.ai caps a single generation at 9 images, 5 MB each. A few clean references that agree with each other still beat a pile that does not.

The Gemini API offers Google Web Search and Image Search grounding for this model, so it can draw on current public information for real places, products and events. The generator on imagesv2.ai currently runs the model without the live-search tool, so it works from the model's own knowledge. Either way, treat any figure, date or claim in a generated image as a draft to verify before publishing.

Nano Banana 2.1 reasons before it draws — that is why layouts, charts and text come out more deliberate than on earlier Flash models. The API exposes three levels (minimal, medium and high; medium is the default) and reasoning cannot be fully disabled. On imagesv2.ai the generator uses Google's default level, which is the balance the model card benchmarks were run at.

Better than any previous Flash-tier model — headlines, menu lines, price tags and short labels come out readable in Latin and CJK scripts alike, and bilingual boards keep their hierarchy. Put the exact wording in quotes in your prompt. Very small or very dense type is still the hardest case for every image model, so proofread anything that ships.

Google's own notes are candid: very small text can still blur, complex spatial relationships (left of, behind, between) occasionally come out wrong, character consistency is strong but not 100%, and the model can still hallucinate details. Keep prompts specific, use references for anything that must match, and generate a couple of variants when the stakes are high.

Google has marked Nano Banana 2 (gemini-3.1-flash-image) for migration and recommends 2.1 for new work. It is still available in the model picker on this page and on the Nano Banana 2 page while it remains in the API, so existing prompts keep working — but 2.1 is cheaper and scores higher on every benchmark, so there is little reason to stay.

No. imagesv2.ai calls gemini-nano-banana-2.1 for you — no Google account, no AI Studio project, no API key, no billing setup. Sign in, write the prompt, press Generate. Everything you make stays in your history.

Yes — what you generate here is yours to use commercially: ads, listings, social content, printed material, merch. You need the rights to anything you upload as a reference, including people's likenesses. Google embeds an invisible SynthID marker in every Gemini image to flag it as AI-generated; it does not change how the image looks.

Images generated on free credits download with a tiled imagesv2.ai watermark. Any purchase — monthly plan, yearly plan or a one-time credit pack — unlocks watermark-free downloads on your account, including for images you generated earlier.
Start generating with Nano Banana 2.1

Open the generator above, claim your free credits, and try Nano Banana 2.1 online. Nano Banana 2, Nano Banana Pro and GPT Image 2 are in the same picker whenever you want to compare.

Looking for a different tool?

Browse every AI image generator, model, converter, and photo tool in one place.

View all AI image tools