$1M AI Film AwardLearn more
A Paris bakery at dawn: gold lettering on the window reads Boulangerie Margaux, Pain au levain, Croissants, depuis 1931, while a baker lays baguettes in the display.
Now live on Segmind

Nano Banana 2.1. Clear text. Faces that hold.

Google's image model for making and editing images: legible text and infographics, characters that stay the same from scene to scene, up to 14 reference images and 1K to 4K. From $0.045 per image.

Try Nano Banana 2.1Nano Banana 2.1 API

Nano Banana 2.1 is Google's image generation and editing model, generally available since 6 October 2026 as the update to Nano Banana 2. It renders legible text and infographics, keeps characters consistent from image to image, takes up to 14 reference images and outputs 1K, 2K or 4K. On Segmind each image costs $0.045 to $0.19, set by resolution, thinking level and web search (as of 31 Oct 2026).

Text and infographics

Labels that land. Diagrams, recipes, posters.

Google lists text rendering among the main improvements in Nano Banana 2.1. Each of these came back with every quoted line spelled and placed as asked, from a nine-label diagram to a five-step method.

A textbook infographic titled Anatomy of a honeybee, with a side view of a worker bee and nine labels: antenna, compound eye, head, thorax, forewing, hindwing, abdomen, stinger and pollen basket.
A labelled diagram
Prompt and settings$0.075 · 39 s

A clean scientific infographic poster on warm off-white paper titled 'ANATOMY OF A HONEYBEE' in bold black capitals. A large, accurate side-view illustration of a worker honeybee in the centre. Exactly nine labels, each used once, with a thin leader line from each label to the right body part: 'ANTENNA' to the antenna, 'COMPOUND EYE' to the large eye, 'HEAD' to the head, 'THORAX' to the hairy middle segment, 'FOREWING' to the front wing, 'HINDWING' to the smaller rear wing, 'ABDOMEN' to the striped rear segment, 'STINGER' to the tip of the abdomen, 'POLLEN BASKET' to the hind leg. A small scale bar at the bottom right reads '5 mm'. Muted amber, black and cream palette, textbook illustration style, generous margins. No other text.

  • Nano Banana 2.1
  • 3:4
  • 2K
  • thinking high

This run: $0.075, rendered in 39 s

A recipe card titled Shakshuka for two beside a pan of shakshuka, listing eight ingredients and five numbered method steps, with the footer Serves 2, 30 min.
A dense recipe card
Prompt and settings$0.075 · 30 s

A printed recipe card photographed flat on a wooden kitchen counter, next to a cast-iron pan of shakshuka. The card's title reads 'SHAKSHUKA FOR TWO'. A left column headed 'YOU NEED' lists: '4 eggs', '400 g chopped tomatoes', '1 red pepper', '1 onion', '2 garlic cloves', '1 tsp cumin', '1 tsp smoked paprika', 'Feta and parsley to serve'. A right column headed 'METHOD' lists five numbered steps: '1. Soften the onion and pepper, 8 min', '2. Add the garlic and spices, 1 min', '3. Pour in the tomatoes and simmer, 10 min', '4. Make four wells and crack in the eggs', '5. Cover and cook until just set, 6 min'. A footer reads 'Serves 2 · 30 min'. Clean modern typography, terracotta accents, soft daylight. No logos or brand names.

  • Nano Banana 2.1
  • 3:4
  • 2K
  • thinking high

This run: $0.075, rendered in 30 s

A bus-shelter poster for the Riverside Loop free shuttle, every 12 minutes from 07:00 to 23:00, with a loop route map naming six stops from Harbour to Central Station, plus one unnamed stop marker.
A transit poster
Prompt and settings$0.075 · 37 s

A bold transit-style poster for a free city shuttle, flat graphic design. Title at the top: 'RIVERSIDE LOOP'. Subtitle: 'Free shuttle · every 12 min · 07:00 to 23:00'. A clean loop-shaped route line in teal with six labelled stops in order: 'Harbour', 'Old Market', 'Library', 'Botanic Garden', 'Stadium', 'Central Station', each with a small round stop marker and a simple icon. A river drawn in pale blue runs through the map. Footer: 'Route 7 · Weekends included'. Teal, coral and cream palette, modern geometric sans-serif, poster mounted at a bus shelter. No logos or brand names.

  • Nano Banana 2.1
  • 3:4
  • 2K
  • thinking high

This run: $0.075, rendered in 37 s

1

Quote every line

Put each piece of text in quotes, exactly as it should read: 'SHAKSHUKA FOR TWO', 'Serves 2 · 30 min'.

2

Pin each label to its place

For diagrams, name every label once and say what it points to: THORAX to the hairy middle segment.

3

Think harder for dense layouts

Set thinking_level to high and output_format to PNG for infographics, menus and posters.

Proofread anything you will print. Our first honeybee run labelled THORAX twice; naming each label once, with the part it points to, fixed it on the next run.

Reference images

Fourteen references. Products and people stay themselves.

Pass up to 14 images in image_urls. Google lists up to 10 objects reproduced with high fidelity and up to 4 characters kept consistent, so a product shoot or a recurring character no longer needs a new model for every frame.

Three products, one shot

Three product shots went in as references, numbered in the order they were sent. The prompt named each one by position; the moth, the serif and the 250 g label all carried over.

  1. Reference 1: a speckled cream stoneware mug with a hand-painted navy moth.1
  2. Reference 2: a canvas tote bag printed with a moth and the words Moth & Millstone, Coffee, Est. 2019.2
  3. Reference 3: a kraft coffee bag with a navy label reading Moth & Millstone, House Roast, 250 g.3
The combined shot: the moth mug full of coffee, the Moth & Millstone tote and the coffee bag with spilled beans on a sunlit linen tablecloth, beside a croissant and a newspaper.
Prompt and settings$0.075 · 43 s

A lifestyle flat-lay photograph on a sunlit oatmeal linen tablecloth, shot from directly above. Use the mug from the first image, filled with black coffee, the canvas tote bag from the second image laid flat, and the coffee bag from the third image lying on its side with a few roasted beans spilling out. Keep each product's exact shape, colours, moth illustration and printed text from its reference. Add a folded newspaper with no readable text, a small plate with a croissant and soft window shadows across the scene. No other logos or text.

  • Nano Banana 2.1
  • 3:2
  • 2K
  • thinking high
  • 3 reference images

This run: $0.075, rendered in 43 s

  • Prompt and settings$0.045 · 16 s

    Studio product photograph of a speckled cream stoneware mug with a hand-painted navy moth on its side, a thick rim and a chunky handle, centred on a seamless plain white background with a soft shadow. No text, no logos.

    • Nano Banana 2.1
    • 1:1
    • 1K
    • thinking minimal

    This run: $0.045, rendered in 16 s

  • Prompt and settings$0.045 · 13 s

    Studio product photograph of a natural canvas tote bag hanging flat, screen-printed in navy with a small moth illustration above the words 'MOTH & MILLSTONE' in a classic serif and 'COFFEE · EST. 2019' in small capitals beneath. Centred on a seamless plain white background with a soft shadow. No other text or logos.

    • Nano Banana 2.1
    • 1:1
    • 1K
    • thinking minimal

    This run: $0.045, rendered in 13 s

  • Prompt and settings$0.045 · 13 s

    Studio product photograph of a matte kraft-paper coffee bag with a folded top and a navy front label reading 'MOTH & MILLSTONE' with a small moth icon, then 'HOUSE ROAST' and '250 g' beneath. Centred on a seamless plain white background with a soft shadow. No other text or logos.

    • Nano Banana 2.1
    • 1:1
    • 1K
    • thinking minimal

    This run: $0.045, rendered in 13 s

One face, three new scenes

The lighthouse keeper from the opening image went in as the only reference. Her face, eyes and silver bob held through a harbour morning, a change of clothes and a storm: three runs, $0.225 in total.

Ines, a lighthouse keeper in her sixties with a silver bob and a yellow oilskin coat, on the gallery of a lighthouse at blue hour with the lamp glowing behind her.Reference
The opening image, cropped here
The same woman, Ines, on a harbour wall in her yellow oilskin, mending an orange fishing net with boats moored behind her.
Mending nets at the harbour
Prompt and settings$0.075 · 33 s

The same woman from the reference image, Ines, keeping her face, silver bob, grey-green eyes, weathered skin and the small scar by her eyebrow exactly. She sits on a stone harbour wall on a bright windy morning, mending an orange fishing net across her knees, wearing the same yellow oilskin coat, fishing boats moored behind her. Documentary photograph, 50mm. No text or logos.

  • Nano Banana 2.1
  • 4:5
  • 2K
  • thinking medium
  • 1 reference image

This run: $0.075, rendered in 33 s

Ines at night in a stone cottage, now in a cream cable-knit sweater with reading glasses in her hair, reading beside a wood stove.
Reading by the stove
Prompt and settings$0.075 · 32 s

The same woman from the reference image, Ines, keeping her face, silver bob, grey-green eyes, weathered skin and the small scar by her eyebrow exactly. At night in her small stone cottage she reads a hardback book in a worn armchair beside a glowing wood stove, now wearing a cream cable-knit sweater and reading glasses pushed up into her hair, a mug of tea on the side table. Warm firelight, cinematic photograph, 50mm. No text or logos.

  • Nano Banana 2.1
  • 4:5
  • 2K
  • thinking medium
  • 1 reference image

This run: $0.075, rendered in 32 s

Ines walking a clifftop path in driving rain, holding the hood of her yellow oilskin, with the lighthouse small in the distance.
Walking out in a storm
Prompt and settings$0.075 · 31 s

The same woman from the reference image, Ines, keeping her face, silver bob, grey-green eyes, weathered skin and the small scar by her eyebrow exactly. She walks along a clifftop coastal path in a storm, rain streaking across the frame, one hand holding the hood of her yellow oilskin coat, the white lighthouse small in the distance behind her. Dramatic grey light, cinematic photograph, 35mm. No text or logos.

  • Nano Banana 2.1
  • 4:5
  • 2K
  • thinking medium
  • 1 reference image

This run: $0.075, rendered in 31 s

Image editing

Edit in plain words. The rest stays put.

Give Nano Banana 2.1 a photo and an instruction. It translated a shop sign, swapped a breakfast and furnished an empty room, and kept nearly everything it was not asked to touch. Drag across each image to compare.

The edit: the same bakery and baker, with the window now reading Margaux Bakery, Sourdough, Croissants, since 1931.
Input: the Paris bakery from the first hero image, its window lettered Boulangerie Margaux in French.
InputEdit
Translate a signThe English lines took the same gold serif and arc; the baker, the bread and the street did not move.
Prompt and settings$0.075 · 37 s

Translate the shop window lettering into English and change nothing else. Replace 'BOULANGERIE MARGAUX' with 'MARGAUX BAKERY' in the same gold-leaf serif, size and position, and replace 'Pain au levain · Croissants · depuis 1931' with 'Sourdough · Croissants · since 1931' in the same style. Keep the baker, the bread, the bicycle, the street, the light and the camera angle exactly as they are.

  • Nano Banana 2.1
  • 16:9
  • 2K
  • thinking medium
  • 1 reference image

This run: $0.075, rendered in 37 s

The edit: the same plate now holds blueberry waffles dusted with icing sugar, and the orange juice is an iced coffee.
Input: a plate of pancakes with a glass of orange juice and a jug of syrup on a wooden table.
InputEdit
Swap the foodOnly the two named items changed; plate, cutlery, jug and wood grain stayed put.
Prompt and settings$0.075 · 30 s

Replace the pancakes on the plate with a tall stack of waffles topped with fresh blueberries and a dusting of icing sugar, and change the glass of orange juice into an iced coffee. Keep the plate, cutlery, wooden table and background exactly the same.

  • Nano Banana 2.1
  • 3:2
  • 2K
  • thinking medium
  • 1 reference image

This run: $0.075, rendered in 30 s

The edit: the same room furnished with a linen sofa, a walnut coffee table on a cream rug, a leather lounge chair, a brass arc lamp, a fig tree and two framed prints.
Input: an empty living room with tall sash windows, oak floorboards, a radiator and white walls.
InputEdit
Stage an empty roomWalls, windows, radiator and floor stayed as shot and the furniture takes the window light; the open door on the right came back closed.
Prompt and settings$0.075 · 41 s

Virtually stage this empty living room for a real estate listing. Treat the photo as the source of truth: keep the walls, ceiling, oak floorboards, windows, door, radiator and camera angle exactly as they are, and keep the soft daylight coming from the windows on the left. Furnish it in a warm mid-century modern style: a three-seat sofa in oatmeal linen facing the windows, a walnut oval coffee table on a cream wool rug, a cognac leather lounge chair, a slim brass arc floor lamp, a tall fiddle-leaf fig in a terracotta pot and two framed abstract prints above the sofa. Scale every piece correctly to the room, give each one real contact shadows that match the window light, and leave a clear walkway to the door. Do not repaint, renovate, add windows or change the view outside.

  • Nano Banana 2.1
  • 3:2
  • 2K
  • thinking high
  • 1 reference image

This run: $0.075, rendered in 41 s

How to edit an image

  1. 1

    Send the photo in image_urls

    Any public image URL works. Add more references to bring in a product, a person or a style.

  2. 2

    Name the change, then what stays

    Say exactly what to change, then list what must not move: the lettering style, the plate, the windows, the camera angle.

  3. 3

    Raise thinking for bigger changes

    Medium handles a swap or a translation. Furnishing a whole room went to high, at the same price per image.

An edit costs the same as a new image at that resolution: each of these was $0.075 at 2K. The breakfast and empty-room photos are the reference images on the Nano Banana 2.1 model page; the bakery is the first image on this page.

Sizes and shapes

Banner to billboard. 1:8 to 8:1, up to 4K.

Pick any of 14 aspect ratios between 1:8 and 8:1, at 1K, 2K or 4K. Google lists the wide and panoramic ratios among this update's improvements, so a site banner or a panorama comes back as one continuous scene.

A 4:1 banner of an Atlantic coastline at golden hour, a lighthouse on the right, with the headline Autumn on the coast and Weekend walks, October to December on the left.4:1 at 2K, 4128x1024
Prompt and settings$0.075 · 44 s

An ultrawide website banner: a rugged Atlantic coastline at golden hour, a small white lighthouse on a headland on the far right, waves breaking on dark rocks, warm light across the grass. The left third is calm sea and sky, with the headline 'AUTUMN ON THE COAST' in large white letters and 'Weekend walks · October to December' in smaller type beneath. One continuous scene. No logos or other text.

  • Nano Banana 2.1
  • 4:1
  • 2K
  • thinking medium

This run: $0.075, rendered in 44 s

An 8:1 illustrated panorama of an imaginary riverside city that moves from dawn on the left through midday and sunset to a starry night on the right.8:1 at 2K, 5856x704
Prompt and settings$0.075 · 29 s

One continuous panoramic illustration of an imaginary riverside city skyline from left to right, as time passes from dawn on the far left, through bright midday and golden sunset, to a starry night with lit windows on the far right. Invented architecture only, with no real landmarks: terracotta rooftops, a green-domed observatory, a slender white footbridge, cranes, rooftop gardens and a ferris wheel, with a river running the whole length and small boats. Flat vector illustration with soft gradients, smooth transitions between times of day, no repeated buildings, no text, no logos.

  • Nano Banana 2.1
  • 8:1
  • 2K
  • thinking medium

This run: $0.075, rendered in 29 s

A vintage botanical plate titled Fungi of the autumn woods with twelve labelled mushrooms, from chanterelle and porcini to black trumpet and parasol.
The 4K original is 5504x3072 pixels, shown here scaled down.
Prompt and settings$0.155 · 58 s

A vintage-style botanical illustration plate on aged cream paper titled 'FUNGI OF THE AUTUMN WOODS' in an engraved serif. Twelve mushrooms drawn in fine watercolour and ink, arranged in a neat four-by-three grid, each with a small handwritten-style caption beneath: 'Chanterelle', 'Porcini', 'Morel', 'Fly agaric', 'Oyster', 'Shiitake', 'Enoki', 'Puffball', 'Chicken of the woods', 'Lion's mane', 'Black trumpet', 'Parasol'. Fine gill lines, spore dust and paper grain visible edge to edge. No other text.

  • Nano Banana 2.1
  • 16:9
  • 4K
  • thinking medium

This run: $0.155, rendered in 58 s

A full-resolution crop of the 4K plate: the lion's mane mushroom drawn in fine ink strokes above its handwritten caption.Full-resolution crop
1280x720 pixels cut from the 4K original at full size: the ink strokes and the handwritten caption hold up at print scale. 4K costs $0.155 with medium thinking, about twice a 2K image.

Thinking levels

Let it think before it draws.

Nano Banana 2.1 can reason about a layout before rendering it. thinking_level takes minimal, medium (the default) or high; medium and high cost the same, high just takes longer. One map prompt, one seed, all three levels at 1K:

  • Minimal thinking: the Pine Hollow map with Easy Street labelled twice and a legend that does not match the prompt.
    Minimal$0.045, 16 sLabelled Easy Street twice and rewrote the legend.
  • Medium thinking: the Pine Hollow map with all six runs, both lifts and the legend as written, plus two compass roses.
    Medium$0.055, 20 sEvery run, lift and legend line right; added two compass roses.
  • High thinking: the Pine Hollow map with all six runs and both lifts labelled once and the legend lines as written, under an added Legend heading.
    High$0.055, 48 sEvery run, lift and legend line right, plus a Legend heading, at over twice the wait.
The prompt, the same at every level: 1:1, 1K

A trail map poster for a fictional ski resort titled 'PINE HOLLOW'. An illustrated snowy mountain with six named runs drawn as coloured lines and labelled: 'Easy Street' and 'Hare Glade' in green, 'Old Mill' and 'Fox Run' in blue, 'Larch Bowl' and 'Summit Chute' in black. Two chairlifts drawn as dashed lines labelled 'Lift A' and 'Lift B'. A legend box in the corner reads 'Green: easy', 'Blue: intermediate', 'Black: expert'. Clean illustrated map style, no other text.

One run per level, so treat it as a pointer rather than a benchmark: use minimal for quick drafts without much text, medium by default, and high when every label has to be right.

Nano Banana 2.1 specifications

Inputs, outputs and limits of Nano Banana 2.1 on Segmind, as of 31 Oct 2026.
SpecificationValue
ModelNano Banana 2.1 (nano-banana-2.1 on Segmind, gemini-nano-banana-2.1 at Google)
InputsA text prompt, an optional system_prompt, and up to 14 reference images in image_urls
Reference limitsUp to 10 object images and up to 4 character images, per Google
Resolution1K (the default), 2K or 4K; no 512px option
Aspect ratio1:1 (the default), 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9, 1:4, 4:1, 1:8, 8:1, or auto
Thinking levelMinimal, medium (the default) or high
Web searchOff by default; set web_search to true to ground the image in Google Search
OutputJPG (the default) or PNG; the image alone, or text and image together
SeedAccepted, but Google may ignore it, so a result is not guaranteed to repeat
Safety tolerance1 (strictest) to 6, with 4 as the default
ResponseThe image file itself from /v1, or a request id to poll from /v2 for queued production jobs
ProvenanceAn invisible SynthID watermark in every image, per Google; C2PA content credentials in each PNG

Nano Banana 2.1 at a glance

Developer
Google
Released
Generally available 6 Oct 2026 as gemini-nano-banana-2.1, the update to Nano Banana 2
Modes
Text to image, and editing or combining from up to 14 reference images
Reference images
Up to 14 in one request: up to 10 objects and up to 4 characters, per Google
Output
1K, 2K or 4K; 14 aspect ratios from 1:8 to 8:1, or auto; JPG or PNG
Thinking level
Minimal, medium (the default) or high reasoning before the image renders
Web search
Optional Google Search grounding for current facts, $0.035 more per image on Segmind
Price on Segmind
$0.045 to $0.155 per image without web search, $0.08 to $0.19 with it (as of 31 Oct 2026)
Typical wait
About 41 s (30-day average on Segmind, as of 31 Oct 2026); 13 s to 58 s in our runs
LMArena rank
#5 text to image (26 Oct 2026) and #7 image edit (23 Oct 2026), the top Google model on both
Provenance
An invisible SynthID watermark in every image, per Google; C2PA content credentials in each PNG
API endpoint
POST https://api.segmind.com/v1/nano-banana-2.1 with an x-api-key header

Nano Banana 2.1 vs Nano Banana Pro and Nano Banana 2

LMArena's blind human votes rank Nano Banana 2.1 above both of its siblings for generating and for editing images, and on Segmind it is the cheapest of the three at every resolution.

LMArena scores and ranks: text to image as of 26 Oct 2026, image edit as of 23 Oct 2026. LMArena marks the Nano Banana 2.1 scores preliminary.
ModelText to imageImage edit
Nano Banana 2.1Gemini Nano Banana 2.11325 (#5)1427 (#7)
Nano Banana 2Gemini 3.1 Flash Image, web search1257 (#11)1387 (#15)
Nano Banana ProGemini 3 Pro Image, 2K1246 (#16)1390 (#14)

Sources: LMArena text-to-image leaderboard and image edit leaderboard. Artificial Analysis's text-to-image arena also puts Nano Banana 2.1 (1176) above Nano Banana 2 (1140). Across all labs, GPT Image 2.5 Sunburst leads both LMArena boards; see the GPT Image 2.5 overview.

One prompt on all three models

The same prompt and seed at 1K on each model, every other setting at its default, on 31 Oct 2026. All three spelled the four-line menu right; Nano Banana 2.1 cost about a third of Nano Banana Pro, and Nano Banana 2 returned fastest.

Prompt: A street food truck parked on a city corner at dusk, painted deep blue with the name 'LA LUNA' in big cream letters on its side. A chalkboard menu propped by the window reads 'TACOS AL PASTOR $4', 'QUESADILLA $6', 'ELOTE $3', 'HORCHATA $3'. A cook in a red bandana hands a paper boat of tacos to a customer, string lights glowing overhead. Photoreal, 35mm. No logos or brand names except the truck name.

The Nano Banana models on Segmind, as of 31 Oct 2026. Prices per image from each model's live pricing page; reference limits from Google's image generation guide.
FeatureNano Banana 2.1Nano Banana 2Nano Banana Pro
Price at 1K$0.045 to $0.055$0.08375$0.1675
Price at 2K$0.065 to $0.075$0.12625$0.1675
Price at 4K$0.145 to $0.155$0.18875$0.30
Web search+$0.035 per image+$0.0175 per image+$0.0175 per image
Resolutions1K, 2K, 4K512px, 1K, 2K, 4K1K, 2K, 4K
Reference images, per Google10 objects, 4 characters10 objects, 4 characters6 objects, 5 characters
Thinking levelMinimal, medium or highMinimal or highNot offered on Segmind
Average wait, 30 daysAbout 41 sAbout 22 sAbout 36 s

Google deprecated Nano Banana 2 on 6 Oct 2026 in favour of Nano Banana 2.1, with no shutdown date announced; it is still available on Segmind. Google keeps Nano Banana Pro as its premium model for the most complex visual tasks.

Where Nano Banana 2.1 fits

Infographics and explainers

Labelled diagrams, recipe cards, maps and step-by-step posters for lessons, docs and social posts, laid out from one prompt.

Product and lifestyle shots

Send the packshots as references and stage them together, with the labels and artwork carried over from each one.

Recurring characters

Storyboards, comics, campaigns and mascots where the same face has to come back scene after scene.

Localised creative

Rewrite the text on a sign, a pack or a poster in another language in an edit, instead of reshooting it.

Real estate and interiors

Stage empty rooms for listings and try furniture or finishes, while the walls, windows and light stay as shot.

Banners and panoramas

Site headers, display ads and wide backdrops at 4:1 or 8:1 that come back as one continuous scene.

Nano Banana 2.1 pricing

Pay per image, with no subscription. Resolution sets most of the price; thinking level and web search adjust it.

Nano Banana 2.1 price per image on Segmind, as of 31 Oct 2026.
ResolutionMinimal thinkingMedium or highWith web search
1K$0.045$0.055$0.08 to $0.09
2K$0.065$0.075$0.10 to $0.11
4K$0.145$0.155$0.18 to $0.19
What common jobs cost at those prices.
JobCost
100 social images at 1K, medium thinking$5.50
100 product shots at 2K, high thinking$7.50
One 4K print master, medium thinking$0.155
One 2K news graphic with web search$0.11
  • One flat price per image: the prompt length, the number of reference images and an edit do not change it.
  • Medium and high thinking cost the same. Minimal saves $0.01 per image.
  • Every Nano Banana 2.1 run on this page billed exactly the price in this table.
Nano Banana 2.1 pricing and calculatorCompare Segmind plans

Nano Banana 2.1 FAQ

What is Nano Banana 2.1?

Nano Banana 2.1 is Google's image generation and editing model, generally available since 6 October 2026 as the update to Nano Banana 2. Its API id at Google is gemini-nano-banana-2.1. It renders legible text and infographics, keeps characters consistent, takes up to 14 reference images and outputs 1K, 2K or 4K images.

How much does Nano Banana 2.1 cost on Segmind?

Nano Banana 2.1 is priced per image on Segmind: $0.045 at 1K with minimal thinking, $0.055 at 1K, $0.075 at 2K and $0.155 at 4K with medium or high thinking. Web search grounding adds $0.035 per image. Every run on this page billed exactly its list price (as of 31 Oct 2026).

Is Nano Banana 2.1 better than Nano Banana Pro?

On LMArena, Nano Banana 2.1 scores 1325 for text to image against 1246 for Nano Banana Pro at 2K, and 1427 for image edit against 1390 (October 2026; 2.1's scores are preliminary). It also costs less: $0.055 against $0.1675 for a 1K image on Segmind. Google keeps Pro for the most complex tasks, with 5 character references to 2.1's 4.

How is Nano Banana 2.1 different from Nano Banana 2?

Nano Banana 2.1 replaces Nano Banana 2, which Google deprecated on 6 October 2026. It ranks higher on LMArena (1325 against 1257 for Nano Banana 2 with web search) and costs less on Segmind: $0.055 against $0.08375 at 1K, $0.075 against $0.12625 at 2K. Nano Banana 2 is faster on average and still offers 512px.

How many reference images can Nano Banana 2.1 use?

Nano Banana 2.1 accepts up to 14 reference images in image_urls. Google lists up to 10 object images kept with high fidelity and up to 4 character images kept consistent. Refer to each one by position in the prompt, such as the mug from the first image, and say what must stay the same.

What do Nano Banana 2.1's thinking levels do?

Nano Banana 2.1 can reason before it renders: minimal, medium (the default) or high. On Segmind minimal costs $0.045 at 1K and medium or high $0.055. In our ski-map test, minimal repeated a trail name, while medium and high labelled all six runs correctly; high took 47 s against 20 s for medium.

Can Nano Banana 2.1 make infographics with readable text?

Yes. Google lists text rendering among Nano Banana 2.1's main improvements, and the diagram, recipe card and shuttle poster on this page came out with every quoted line correct. Put each line in quotes, use thinking level high and PNG for dense layouts, and proofread: our first honeybee run labelled THORAX twice.

What resolutions and aspect ratios does Nano Banana 2.1 support?

Nano Banana 2.1 outputs 1K, 2K or 4K, with no 512px option. It offers 14 aspect ratios from 1:8 to 8:1, including 4:1 and 8:1 panoramas, or auto. In our runs a 16:9 image came back at 2752x1536 at 2K and 5504x3072 at 4K.

How do I edit a photo with Nano Banana 2.1?

Send the photo in image_urls, describe the change, then name what must stay the same. On this page Nano Banana 2.1 translated a shop sign from French to English, swapped pancakes for waffles and furnished an empty room, and each photo kept its layout and light. Use thinking level high for bigger changes.

What does web search do in Nano Banana 2.1?

With web_search set to true, Nano Banana 2.1 grounds the image in Google Search results, which helps with news, live data and recent products. It adds $0.035 per image on Segmind. Leave it off for creative work that does not need current facts.

Are Nano Banana 2.1 images watermarked?

Yes. Google states that every Nano Banana 2.1 image includes SynthID, an invisible watermark that marks it as AI generated. The PNG files we downloaded from Segmind also carried C2PA content credentials signed by Google, recording the generation and the SynthID watermark. Converting or re-saving a file can strip the C2PA metadata, so keep the original PNG where provenance matters.

How do I use the Nano Banana 2.1 API?

Create a Segmind API key, then POST a prompt to https://api.segmind.com/v1/nano-banana-2.1 with the key in the x-api-key header; add image_urls to edit or combine. Nano Banana 2.1 returns the image file in the response. The model page has curl, Python and JavaScript examples and every parameter.

Your next infographic, product shot or edit is one request away.

One API key for Nano Banana 2.1, from $0.045 an image.

Try Nano Banana 2.1View API docs

Last updated