Nano Banana Prompt Guide: How to Write Prompts That Give You Usable Images

Summary

Learn how to write better Nano Banana prompts for Google’s Gemini image models using clear formulas, practical examples, and editing techniques. This guide explains how to structure prompts for product images, portraits, infographics, readable text, and consistent characters, while avoiding common mistakes like vague keywords, missing aspect ratios, and confusing reference-image instructions.

‍

Nano Banana is the nickname for Google's Gemini image models, the ones behind nearly every nano banana prompt guide search that's popped up since 2025. If you've typed a vague line into Gemini and gotten back a flat, slightly off image, the model usually isn't the problem. The prompt is.

Nano Banana doesn't read prompts the way older image generators do. It reasons through what you're asking before it draws anything, sometimes called a Gemini Nano Banana prompt since the whole system runs on Google's Gemini architecture. That means clear, structured language gets you far better results than a string of style keywords ever will.

This guide breaks down how to write prompts for Nano Banana, Nano Banana 2, and Nano Banana Pro, with copy-ready Nano Banana prompt examples for generation, editing, text, product shots, and character consistency. Whether you're a marketer building ad creative, a designer testing concepts, or a video editor pulling keyframes, the same core rules apply.

What is a good Nano Banana prompt?

A good Nano Banana prompt formula reads like a creative brief, not a list of tags: 

Subject + Action + Setting + Style + Composition + Lighting + Text + Constraints.

State what should change and what should stay the same when you're editing, write in full sentences instead of comma-separated keywords, and put your most important detail first, since Nano Banana weighs earlier words more heavily than later ones. 

For edits, attach a reference image and describe the change directly instead of starting over.

If you only take one rule from this guide: write Nano Banana prompts like you're briefing a photographer, not typing search terms

What is Nano Banana?

Nano Banana is the public nickname for Google's Gemini image generation and editing models. There isn't just one. As of mid-2026, there are three:

• Nano Banana, running on Gemini 2.5 Flash Image, the original model that went viral in 2025.
• Nano Banana 2, running on Gemini 3.1 Flash Image, is now the default image model inside the Gemini app.
• Nano Banana Pro, running on Gemini 3 Pro Image, the higher-fidelity option for text-heavy or complex work.

All three sit on Gemini's reasoning architecture instead of a traditional diffusion pipeline, and that's worth understanding before you write a single prompt. 

Diffusion models like Midjourney or Stable Diffusion match keywords to visual patterns. Nano Banana processes a prompt closer to how it processes a conversation, which is why it understands relationships and intent instead of just nouns and adjectives.

If you've prompted Google's older Imagen models before, expect an adjustment period. Imagen migration to Nano Banana isn't a like-for-like swap. Nano Banana rewards narrative, descriptive prompts, and conversational follow-up edits, where Imagen leaned more on shorter structured tags.

In practice, this means you can prompt Nano Banana the way you'd brief a person, including quick follow-ups like "make it warmer" or "move the subject slightly left," and it understands what you mean without you restating the whole scene.

Nano Banana also pulls in real-world knowledge through the Gemini API, and on the newer models, live web search. That's why it can label a real landmark correctly or build a roughly accurate infographic about a recent event, something most image generators simply can't do.

Every image Nano Banana produces carries SynthID, Google's invisible watermark that marks the file as AI-generated. It doesn't change how you prompt, but it matters if you're publishing the images professionally and someone asks.

You can reach Nano Banana through the Gemini app, Google AI Studio, Google Search's AI Mode, or the Gemini API for developers building it into their own products.

Nano Banana vs. Nano Banana 2 vs. Nano Banana Pro: Which one should you use

This is the most common point of confusion, and the answer changed recently. In February 2026, Google made Nano Banana 2 the default image model across the Gemini app, replacing Nano Banana Pro in that role. 

Here's how to think about the three.

Nano Banana, on Gemini 2.5 Flash Image, is the original. It's fast and cheap, fine for casual social content or rough drafts, but it's mostly been succeeded by the newer two.

Nano Banana 2, on Gemini 3.1 Flash Image, is the current default. It's built for speed without giving up much quality, handles character consistency across multiple images reasonably well, and picked up live search grounding that used to be Pro-only. For most day-to-day work, start here.

Nano Banana Pro, on Gemini 3 Pro Image, is the specialist. It costs more and runs slower, but it's still the better choice when a project has heavy text, complex layouts like infographics, or needs the highest fidelity for print or a hero campaign image. Most of the Nano Banana 2 prompts in this guide will also work in Pro, just slower and with sharper output.

A simple way to decide: 

• default to Nano Banana 2 for nearly everything
• Switch to Pro only when the output needs precise typography, dense layout logic, or you're producing a final asset where a redo costs more time than Pro's extra processing does. 

If you're writing Nano Banana Pro prompts specifically, lean on the same formula and add more explicit typography and layout detail, since that's where Pro earns its keep.

How to write better Nano Banana prompts

Once the formula is down, the gap between an okay image and a genuinely usable one comes down to specificity.

• Talk like a creative director, not a search engine. Instead of "professional photo," try "shot on an 85mm lens at f/1.8, soft three-point studio lighting, shallow depth of field." Nano Banana understands photography and cinematography language because it's trained on real-world knowledge, not just art tags.
• Describe lighting on purpose. A few phrases worth keeping handy: golden hour backlighting for long shadows, overcast daylight for even soft tones, harsh studio lighting for high contrast. Lighting language does more for realism than almost any other single detail.
• Write constraints as positive statements. Nano Banana has no negative prompt field, the way some other tools do. You can't list "no blur, no extra fingers" in a separate box. Fold exclusions into the main prompt as direct, positive statements instead: natural hand anatomy with five fingers, sharp focus throughout, clean background with no added text.
• Build a character or brand card for repeat use. If you're generating several images that need to match, write a short, fixed description once, covering appearance, outfit, color palette, accessories, and reuse that exact wording across every prompt. This one habit prevents more drift than any other fix on this list.
• Keep first, change second, when editing. For portraits and edits, name what to preserve before naming what changes. "Keep this woman's face and hair exactly as shown, change her outfit to a navy blazer" works better than the reverse order.
• Iterate inside one conversation instead of starting fresh chats. Nano Banana holds onto context from earlier messages, which is part of why small follow-up edits work so well. Opening a new chat for every tweak means re-explaining details you'd already established.

The Nano Banana prompt formula

There isn't one Nano Banana prompt formula. There are two, and most guides only give you one of them. We will show you both!

Formula 1: Generating an image from scratch

Subject + Action + Setting + Style or medium + Composition or camera + Lighting and color + Text, if the image needs words + Constraints

This is the core Nano Banana image generation prompt structure, and it covers most requests on its own. 

Example: "A barista in a denim apron steaming milk behind a marble counter, in a sunlit corner coffee shop. Photorealistic, shot on a 50mm lens, soft window light from the left. No extra people in frame."

Put your most important detail first. Nano Banana weighs earlier words more heavily than later ones, so if lighting matters more to you than the background, lead with the lighting.

Formula 2: Editing or combining reference images

This second formula exists because editing needs a different mindset than generating from nothing. You already have a starting image, so your prompt's job is to state what changes and what stays exactly as it is.

Reference image + Relationship instruction + New scenario

Example: "Using the attached product photo as the reference, place this bottle on a rustic wooden table with morning light coming through a window, keeping the label and proportions exactly as shown."

A few rules apply to both formulas:

• Write full sentences, not comma-separated keyword lists. Nano Banana reasons through language; it doesn't just pattern-match tags.
• Avoid contradictions inside the same prompt, like asking for a minimal background and a busy, detailed background at once.
• If your results start drifting from what you asked for, simplify. Break long sentences into shorter clauses.

25 Copy-paste Nano Banana prompts to try

In this section, we will cover the use cases that creators and marketers ask for most. Swap the bracketed details for your own subject and keep the rest of the structure intact.

Consider this section your working set of best Nano Banana prompts to start from, rather than building from zero.

Product and e-commerce

1. A matte ceramic coffee mug on a marble countertop, soft morning light from a north-facing window, photorealistic product photography, shallow depth of field, no text.
2. A pair of white running shoes on a concrete platform, studio lighting from two sides, a clean grey background, and a commercial e-commerce style.
3. A glass skincare bottle with a pump top on a wet slate surface with condensation droplets, soft diffused lighting, and a premium cosmetics aesthetic.
4. A leather handbag on a marble pedestal, side lighting with a soft shadow, editorial luxury retail style, no visible logos.
5. A smartwatch on a minimalist white display stand, gradient backdrop, clean studio shadows, modern tech product photography.

Portraits and people

6. A professional headshot of a man in a navy suit, shot on an 85mm lens, three-point studio lighting, neutral grey background, natural skin texture with visible pores, not airbrushed.
7. A candid portrait of a woman laughing outdoors, golden hour light, shallow depth of field, photojournalistic style, no posed expression.
8. A close-up portrait with cinematic lighting, half her face in shadow, the other half lit by a single warm key light, moody color grade.

Marketing and social content

9. A flat lay of a coffee bag, scattered beans, and a wooden spoon on a rustic table, warm lifestyle photography, soft natural light, social media ready.
10. A bold Instagram post graphic announcing a "Summer Sale," bright yellow background, large sans-serif headline text, minimal layout, no clutter.
11. A YouTube thumbnail with a surprised expression on the subject's face, bright saturated colors, bold white outlined text reading "You Won't Believe This," and high contrast.
12. A flat, modern illustration of a small team brainstorming around a table, friendly color palette, clean vector style, for a blog header.

Scenes and environments

13. A cozy reading nook by a window, rain visible outside, warm lamp light, soft shadows, and an illustrative but realistic style.
14. An aerial view of a winding coastal road at sunset, cinematic color grade, wide-angle perspective, no people or cars.
15. A minimalist modern kitchen with morning light streaming across the floor, architectural photography style, no people.

Text and typography

16. A motivational poster with bold black text reading "KEEP GOING" centered on a soft gradient background, clean sans-serif font, and generous white space.
17. A magazine cover with the title "CREATIVE FUTURE" in bold white sans-serif text at the top, positioned behind the subject's head but fully legible.
18. A storefront sign mockup reading "OPEN DAILY 8 AM TO 6 PM" in clean black lettering on a white wooden sign, soft daylight.

Brand and design

19. A simple logo concept for a coffee brand called "Northbound," minimal line art of a compass, single color, clean and modern.
20. A business card mockup for a design studio, soft shadows, neutral background, modern sans-serif typography, no placeholder text.
21. A color palette mockup card showing five swatches labeled "Sunset Collection," a clean grid layout, and soft drop shadows.

Creative and conceptual

22. A surreal image of a tiny house balanced on top of a giant teacup, soft pastel lighting, and whimsical illustration style.
23. A double exposure portrait blending a forest silhouette into a person's profile, with moody black and white tones.
24. A vintage travel poster style image of a lighthouse on a cliff, muted retro color palette, bold, simple shapes, 1960s travel ad aesthetic.
25. A 3D isometric illustration of a cozy home office desk setup, soft pastel colors, clean rounded shapes, and subtle shadows.

Nano Banana image editing prompts

Editing is where Nano Banana separates itself from most generators, because you can talk to it conversationally instead of starting over every time.

The golden rule: if an image is close, don't regenerate from scratch. Ask for the specific change instead.

A few Nano Banana image editing prompts worth keeping on hand:

• Remove the red car in the background and replace it with a grey sedan that matches the lighting of the street.
• Change the time of day to golden hour. Add long shadows from the left and warm the highlights without changing anything else in the scene.
• Replace the text on the sign with "GRAND OPENING" in the same font style and color, keeping the rest of the image unchanged.
• Keep this person's face and pose exactly the same, and change only the jacket color to forest green.
• Combine these two images: use the lighting and background from the first image, and place the product from the second image into the scene.

When you're working with more than one reference image, name them by what they are instead of by position. Say "the wireframe sketch" and "the finished render," not "image 1" and "image 2." 

Nano Banana sometimes treats a positional reference as an instruction to reproduce that exact image, which is the opposite of what you wanted.

Stick to a handful of high-quality reference images rather than uploading a dozen. More references don't automatically mean more accuracy. Overloading a prompt with images can hurt consistency instead of helping it.

If you're batching several similar edits, like swapping a product across ten lifestyle shots, keep the editing instruction worded identically each time and only change the product reference. Consistent phrasing produces more consistent results across the batch.

Nano Banana prompts for infographics and diagrams

This is one of the more underused capabilities of Nano Banana Pro and Nano Banana 2, especially for marketers and small teams who need a quick visual explainer without opening a design tool.

• For step-by-step processes, ask for an S-curve or zigzag layout so the eye follows a natural sequence. For broader overviews, request a bento grid, a layout of clean rectangular sections, one per topic. 
• For anything meant to be scanned quickly, specify a three-level text hierarchy: headline, subheader, short body text, plus enough white space that it doesn't feel cramped.

Example: "Create a clean, modern vertical timeline infographic showing five product milestones from 2022 to 2026. White background, thin grey dividers, circular markers for each year, minimal icons, short text blocks, bold header reading 'Our Journey.'"

Example: "Generate a bento grid infographic titled 'Q2 Marketing Recap.' Use four rectangular sections, each with an icon, a bold title, and two to three short bullet points. Pastel color palette, clean modern typography, subtle drop shadows."

Because Nano Banana Pro and Nano Banana 2 can pull in real-world context through Google Search, you can also ask for current information inside an infographic. That said, always fact-check anything time-sensitive before you publish it. The model is good, not infallible.

This approach works particularly well for internal reporting decks or quick social graphics, the kind of visual where waiting on a designer isn't worth the turnaround time.

How to get clear text in Nano Banana images

Text rendering used to be the weak point of nearly every image generator. Nano Banana is one of the first to make it reliably legible, but you still need to prompt it correctly. Here are some tips for the best Nano Banana prompts:

• Put your exact text in quotation marks so the model treats it as a literal string, not a description. "A poster with the headline 'NEW ARRIVALS' in bold red letters works better than describing the words without quotes around them.
• Keep the text short. Long sentences or full paragraphs inside an image are far more likely to come out garbled. If you need more than a headline and one line of supporting copy, generate the graphic without text and add the copy in a design tool afterward.
• Specify font style and placement. "Bold sans-serif font, centered at the top, white text with a thin black outline" gives the model far more to work with than just "add a title."

If you need text in another language, Nano Banana can translate and localize copy inside an image while keeping the original layout, lighting, and style intact. That's useful for marketing teams localizing one creative across several markets. Keep translated text short for the same reason as before; shorter strings render more reliably, and have a native speaker review anything going out for actual publication.

If a logo or wordmark needs to stay pixel-perfect across multiple images, generate it once as a clean reference and ask Nano Banana to place that exact reference into new scenes rather than redrawing the text from scratch each time.

How to keep characters consistent in your Nano Banana images

Character drift, where the same character looks slightly different from one image to the next, is the most common complaint about AI image tools. Nano Banana handles it better than most, if you set it up correctly.

• The most reliable method is building a character reference sheet before generating anything else. In a single session, generate two or three images of your character from different angles: looking left, looking right, and straight on. Use those images as references for every future prompt involving that character.
• Write a short, fixed character description once and reuse that exact wording every time. Something like: "A woman in her late twenties, sharp green eyes, shoulder-length auburn hair, wearing a brown leather jacket." Don't rewrite this from scratch for each new prompt. Copy and paste it.
• For scenes with more than one character, map their positions explicitly. "Character A on the left, Character B in the center, Character C on the right" stops the model from mixing up who's who.
• If you're building a sequence, like a comic strip or storyboard, reference the previous image directly in your next prompt. "Same character from the previous shot, now shown in a close-up with a worried expression," keeps continuity across panels without re-describing everything from scratch.

Where can you access Nano Banana?

You don't need to hunt for a separate app. Nano Banana models are built into:

• Pixelbin; the most efficient workspace for creators and e-commerce teams. It combines Nano Banana image generation and advanced editing in one streamlined interface. 
• The Gemini app, for everyday generation and editing.
• Google AI Studio, for testing prompts and comparing Nano Banana 2 against the Pro side by side.
• Google Search's AI Mode, for quick visual answers inside search.
• The Gemini API is for developers building Nano Banana into their own apps or tools.

A growing list of third-party design and video platforms also routes requests to these same underlying models. If you're already working inside one of those tools, there's a decent chance you've been using Nano Banana without realizing it.

Common Nano Banana prompt mistakes you must avoid

Most disappointing results trace back to one of these habits. Below are some of the common Nano Banana prompt mistakes: 

• Writing tag soup instead of sentences: "Woman, park, sunny, 4k, realistic" gives Nano Banana far less to reason through than an actual sentence describing the scene.
• Stacking contradictions: Asking for a minimal clean background and a busy detailed background in the same prompt forces the model to guess which one you meant.
• Assuming there's a negative prompt field: There isn't. Exclusions need to be written as positive instructions inside the main prompt.
• Referencing images by position instead of description: "Image 2" is ambiguous. "The product photo" or "the sketch" isn't.
• Regenerating instead of editing: If an image is mostly right, asking for a fresh generation usually loses the parts that were already working. Ask for the specific change instead.
• Overloading one prompt with unrelated instructions: If you're asking for a new subject, a new background, new lighting, and new text all at once, split it into two prompts.
• Forgetting to lock the aspect ratio: Without it, Nano Banana defaults to a square image even if your final use case needs landscape or vertical.

None of these mistakes takes much to fix. Most just need one extra sentence of thought before you hit generate.

Troubleshooting: What to fix when the output looks wrong

1. The aspect ratio is wrong. Nano Banana defaults to a 1:1 square unless told otherwise. State the ratio explicitly at the end of your prompt, like "16:9 aspect ratio," and if you're in the API or AI Studio, set it in the generation settings rather than relying on the prompt text alone. If it still won't comply, attach a blank reference image shaped like the frame you want and ask the model to fill it.

2. The Pixelbin Advantage: When generating images inside Pixelbin, you can skip text-prompt hacks entirely. You can select your exact desired canvas dimensions, like 16:9 landscape or 9:16 vertical, directly from the visual settings menu before typing your prompt. The platform handles the model constraints automatically to deliver the correct shape.

3. The image looks blurry or low-res. Add a direct quality instruction like "sharp focus, high resolution, fine detail." If that doesn't fix it, request a specific output size like 2K or 4K instead of leaving resolution unspecified. If you are generating images with Pixelbin, you simply try the image upscaler after generating the output. You can enhance images by up to 8X while preserving quality. 

4. You tried to use a negative prompt and hit an error. Nano Banana's API doesn't support a separate negative prompt field. Move whatever you wanted excluded into the main prompt as a positive statement instead.

5. The model returned your original image unchanged. This usually happens with multi-image edits where the wording got read as "reproduce this image" instead of "use this as a reference." Rewrite the prompt with semantic names for each image and lead with a clear action verb like replace, combine, or relight.

6. Hands or faces look distorted. Add explicit anatomy instructions: natural hand anatomy with five fingers, symmetrical facial features. Regenerating a couple of times helps too, since output quality varies run to run even with an identical prompt.

7. Quality degrades after several rounds of edits. Each edit compounds on the last one. If an image has gone through five or six rounds of changes and looks worse than where it started, go back to an earlier version and make fewer, larger changes instead of many small ones.

8. Colors look oversaturated or off. Name a specific color grade instead of leaving tone to chance, like "muted, true-to-life color, not oversaturated." Nano Banana tends to punch up saturation by default unless told otherwise.

How Nano Banana compares to other AI image tools

Nano Banana's biggest edge over Midjourney is text rendering and conversational editing. Midjourney still produces strong artistic and stylized imagery, but clear enough in-image text and precise follow-up edits aren't its strong suit. Nano Banana, especially the Pro version, handles both well.

Compared to GPT Image, the two sit closer together, particularly in terms of text accuracy. The practical difference often comes down to the ecosystem. If your workflow already runs through Google Cloud, Vertex AI, or the Gemini API, staying inside Nano Banana avoids extra integration work.

Compared to Adobe Firefly, Nano Banana tends to produce stronger photorealism, while Firefly has the advantage of sitting directly inside Photoshop and Premiere for teams already working in Adobe's tools.

There's no universal winner. The right choice depends on whether your priority is artistic style, text accuracy, photorealism, or how well a tool fits into what you already use day to day.

Cost and access are worth weighing, too. Nano Banana 2 generally costs less per image than Nano Banana Pro. Rates change as the models update, so always refer to Google's current pricing page.

Nano Banana Prompt Checklist

Before you publish anything built from this nano banana prompt guide, run through this:

• Is the subject and action stated in a full sentence, not a list of tags?
• Have you included setting, style, and lighting?
• If there's text in the image, is it in quotation marks with a font and placement noted?
• Have you stated the aspect ratio explicitly?
• Are exclusions written as positive statements instead of a negative prompt?
• If you're editing, have you said exactly what to keep and what to change?
• If a character needs to stay consistent, are you reusing the same description and the same reference images?
• Are you editing an existing image instead of regenerating from scratch when the result is already close?

Conclusion

Getting great results with Nano Banana comes down to clear communication. Whether you're creating images from scratch or editing existing ones, well-structured prompts help you achieve better results with less trial and error. The Nano Banana prompts copy paste examples in this guide give you a strong starting point, but the real improvement comes from adapting them to your own goals. Keep your instructions simple, be specific about what you want, and refine your prompts as you go. With practice, Nano Banana becomes a reliable creative tool for producing high-quality images faster and more consistently.