Nano Banana Prompts: 15 Copy-Ready Prompts for 2.1
Fifteen copy-ready Nano Banana prompts for Google's Nano Banana 2.1, grouped by job: portraits and products, surgical edits, posters and infographics, 14-image composition and 8:1 banners. Plus when to switch to Pro or Lite.
By Eftal Yurtseven · Nov 17, 2026 · 9 min read

Most Nano Banana prompts fail the same way. The prompt reads like a pile of tags ("cyberpunk, neon, 4k, cinematic, cat"), the model guesses the rest, and you spend ten runs undoing that guess. Good nano banana prompts read like a brief to a photographer who also happens to be a typesetter and a retoucher.
That brief got more rewarding on October 6, 2026, when Google released Nano Banana 2.1. Google's DeepMind model card says it's built on Gemini 3.6 Flash, and the official Gemini API docs list what changed from Nano Banana 2: better visual quality at 1K, 2K and 4K, better prompt adherence, steadier characters across multiple turns, sharper text rendering and infographic layout, and fixed tiling artifacts on the very wide and very tall frames. Every one of those upgrades rewards a better prompt, not a longer one. Below are 15 copy-ready prompts grouped by job, plus when to switch to Nano Banana Pro or Nano Banana 2 Lite.
Pick the Right Banana Before You Write a Word
Nano Banana 2.1 is Google's recommended default now. It keeps Flash speed, takes up to 14 reference images, supports 14 aspect ratios including 1:4, 4:1, 1:8 and 8:1, and on each::labs exposes a thinking_level switch: "minimal" for fast drafts, "high" when a prompt has a lot of moving parts. For editing, the Nano Banana 2.1 Edit endpoint accepts 1 to 14 input images, and setting the ratio to Auto keeps the shape of your source photo.
Nano Banana Pro (Gemini 3 Pro Image) is the one Google positions as its premium model for complex tasks: intricate instructions, brand consistency, detailed text. If your prompt reads like a creative director's spec sheet, Pro earns its slower turn. Its edit sibling, Nano Banana Pro Edit, takes up to 10 input images.
Nano Banana 2 Lite is the fastest of the three and built for volume: thumbnails, variant grids, first drafts. Google says it isn't optimized for multiple references or multi-turn editing, so give it single, self-contained prompts and keep the stacking for 2.1. If you want the long version of this decision, our Nano Banana 2 vs Nano Banana Pro comparison walks through it.

The Anatomy of Nano Banana Prompts That Work
Google's own prompting guidance boils down to one line: describe the scene, don't list keywords. Relationships between things are what the model reasons about. "A cat on a windowsill" is a subject. "A ginger cat asleep on a rain streaked windowsill" is a moment.
Build that moment in a fixed order and you'll rarely get lost. Start with the shot type and the subject, then the setting, then the light, then the camera and lens, and finish with the purpose. That last part is the one people skip, and Google's docs say stating who the image is for (a menu, a book cover, a landing page) beats a bare request. Intent decides a hundred small choices you'd never write down.
Phrase things positively: "an empty cobblestone street at dawn," not "a street with no cars," because naming the unwanted thing often plants it. And use real photography words. "85mm portrait lens, shallow depth of field" or "slightly elevated 45 degree product shot" carries more information than "beautiful" ever will.

Text-to-Image Nano Banana Prompts for 2.1
These four cover the jobs most people open the text-to-image tab for. Each one names a purpose in its last sentence.
1. The editorial portrait. Lighting and lens do the heavy lifting.
A close-up portrait of a 60-year-old fisherman with a salt-and-pepper beard, wearing a heavy navy wool sweater, standing on a weathered wooden pier at golden hour. Warm low sun from the left rakes across the deep lines of his face; the harbor behind him falls into soft bokeh. Shot on an 85mm lens at f/1.8, natural skin texture, muted film color. For the cover of a regional food magazine.
2. The product hero. Materials and surfaces matter more than adjectives.
A high-resolution studio product photo of a matte black ceramic coffee mug on a polished concrete surface. Three-point softbox lighting with a gentle rim light that traces the handle, a faint wisp of steam rising from the coffee. Slightly elevated 45 degree angle, clean warm-grey seamless background, crisp focus on the glaze texture. Square format for an online store listing.
3. The stylized illustration. Google's guide asks you to name the medium and line quality when you leave photography behind.
A children's book illustration of a small fox carrying a lantern through a snowy birch forest at night. Gouache on textured paper, soft rounded shapes, thick confident outlines, a limited palette of deep blue, cream and warm orange. The lantern is the only light source and casts a soft circle on the snow. Leave calm empty space in the upper third for the page text.
4. The step-by-step scene. For busy compositions, build back to front.
Background: a rainy Tokyo side street at night, glowing paper lanterns and a vending machine casting cyan light. Midground: a tiny ramen counter with steam rolling off three bowls. Foreground and focal point: a chef in a white headband laughing as he hands a bowl across the counter, sharp focus on his hands and the bowl. Wide-angle 24mm, wet reflections on the asphalt, cinematic color grade. A still for a travel documentary poster.
Edit Prompts: Change One Thing, Name What Stays
Editing is where lazy prompts do the most damage. The rule: say exactly what changes, then say out loud what must not. Google's template opens with "Using the provided image of..." to mark the ground truth.
5. The targeted swap. Inpainting without a mask, just language.
Using the provided image of the living room, change only the blue sofa into a vintage cognac leather Chesterfield sofa. Keep the rug, the pillows, the plants, the window light and the camera angle exactly as they are. Match the existing shadows and color temperature.
6. The identity-safe restyle. Faces and logos need their protection spelled out.
Using the provided photo of the woman, change her outfit to a tailored olive linen suit with a white t-shirt and swap the background to a sunlit Lisbon street with yellow trams. Her face, hairstyle, eye color, skin tone and expression must remain completely unchanged. Keep the same pose, framing and soft daylight.
7. The style transfer. Keep the composition, trade the medium.
Recreate the provided photograph of the city skyline as a 1950s screen-printed travel poster. Preserve the exact composition, building silhouettes and horizon line. Flat layered color in four inks (teal, coral, mustard, off-white), visible halftone grain, slightly misregistered edges.
After the first edit lands, follow up with one change per turn ("now make it late afternoon, keep everything else the same"). Stronger multi-turn consistency is a stated 2.1 upgrade, and small steps are how you cash it in.

Text in Images: Posters, Labels and Infographics
Clear text for posters and diagrams is an intended use Google calls out on the 2.1 model card. The pattern never changes: quote the exact words, describe the type, describe the layout. Keep the copy short (headlines and labels render far more reliably than paragraphs) and raise the resolution to 2K or 4K when the lettering is small. We covered this in more depth in Nano Banana 2.1: Wider Frames, Sharper Text.
8. The event poster.
A vertical poster for a jazz night. At the top, the headline "BLUE HOUR SESSIONS" in a tall condensed art deco typeface, cream on deep midnight blue. Center: a stylized gold saxophone made of flowing brass curves. At the bottom in a small clean sans serif: "Fridays 9 PM, The Lantern Room". Generous margins, balanced grid, subtle paper texture.
9. The packaging label.
A product photo of a glass jar of wildflower honey on a rustic wooden table. The kraft paper label reads "MEADOW GOLD" in a hand-drawn serif with a small bee illustration, and underneath in smaller letters "Raw Wildflower Honey, 350 g". Soft morning window light, shallow depth of field, honey glowing amber.
10. The infographic. 2.1's infographic layout accuracy is a stated upgrade, so give it a clear structure to fill.
A clean flat-design infographic titled "How Coffee Gets to Your Cup" with five numbered steps in a left-to-right flow: "Harvest", "Process", "Roast", "Grind", "Brew". Each step has a simple icon and a one-line label. White background, a warm brown and terracotta palette, rounded sans serif type, thin connecting arrows. For a café's social media carousel.
Multi-Image Composition With Up to 14 References
This is the job prose can't do alone. Upload the pieces, then refer to them by order, exactly as Google's docs do ("the first image"). Ambiguity about which photo is which is what makes compositions drift. On 2.1 the ceiling is 14 references, with up to 4 characters and 10 objects in one workflow.
11. The virtual try-on.
Take the linen dress from the first image and put it on the model in the second image. Keep the model's face, body proportions and pose from the second image. Match the fabric drape and wrinkles to her pose, and relight the dress to the soft overcast daylight of the second image. Full-length fashion catalog shot.
12. The group scene from separate portraits.
Place the three people from images one, two and three around a dinner table in a warm candlelit Italian trattoria, laughing mid-conversation. Person one sits on the left, person two in the center raising a glass, person three on the right. Keep each face and hairstyle identical to their reference photo. Use the ceramic plates from image four on the table. Eye-level 35mm shot, warm tungsten light.
Spell out positions and anchor them to objects, since left and right are easy to mix up in a crowded frame. If a set like this turns into a repeatable pipeline, chain the steps in each::labs Flows rather than redoing them by hand.

Aspect Ratio Tricks: Banners, Strips and Skyscrapers
The 1:4, 4:1, 1:8 and 8:1 ratios aren't brand new (Nano Banana 2 had them too), but 2.1 is where Google fixed the tiling artifacts that used to creep into those extreme frames at 2K and 4K. Compose for the shape: an 8:1 strip is read like a sentence, left to right, so tell the model where the eye travels.
13. The 8:1 website banner.
An ultra-wide panoramic banner of a misty alpine valley at sunrise. From left to right: a dark pine forest, a winding silver river, a small red cabin with smoke rising from the chimney at the right third, then soft peaks fading into pink haze. Keep the center calm and low in detail so headline text can sit on top. Painterly realism, soft atmospheric perspective.
14. The 1:4 vertical scroll.
A tall vertical illustration that reads top to bottom: a starry night sky at the top, a lighthouse beam cutting through clouds, the lighthouse on a cliff in the middle, waves crashing on rocks below, and a tiny rowboat with a lantern at the very bottom. Muted Prussian blue and warm yellow palette, woodcut texture. For a long-scroll mobile story.
15. The 4:1 product lineup.
A wide studio shot of six sneakers lined up in a single row on a pale sage background, each a different colorway from cream to charcoal, evenly spaced with identical soft shadows. Eye-level, even diffuse lighting, crisp detail on the stitching. A header image for a collection page.
Set the shape in the aspect_ratio field on Nano Banana 2.1 Text to Image, and let the prompt describe the composition that fits it.
Iterate Like an Editor, Not a Gambler
Great images come from a decent first prompt and small, deliberate corrections. Lock the seed while you tweak wording so you can see what each change actually did, and use num_images for up to four takes while you're still choosing a direction.
Switch thinking_level to "high" for prompts with many constraints (like 4, 10 and 12) and "minimal" for quick sketches. When a draft is wrong in three places, fix one, look, then fix the next. For a look at how these patterns hold up in practice, see our Nano Banana 2.1 review.
Frequently Asked Questions
What makes good nano banana prompts?
Sentences beat keyword lists. Describe subject, setting, light and camera, then say who the image is for. Quote any text exactly, and when editing, list what must stay unchanged.
Should I use Nano Banana 2.1 or Nano Banana Pro?
Start on 2.1, which Google now recommends as the default. Move to Nano Banana Pro when a prompt is dense with brand rules and typography and the extra reasoning is worth the wait.
Can Nano Banana 2.1 render text in other languages?
Yes. Localized text across several languages is one of the uses Google lists on the model card. Quote the words in the target language and keep them short.
How many reference images can I give Nano Banana 2.1?
Fourteen on the 2.1 Edit endpoint, with up to 4 characters and 10 objects in one workflow according to Google. Name each image by its order in the prompt.



