Nano Banana Prompts

120+ copy-paste prompts for Nano Banana 2 and Nano Banana Pro, organised around portraits, edits, products, readable text, pets, family photos, infographics and repeatable visual series. Paste, attach a reference when needed, and inspect the result.

Last updated September 22, 2026

What Nano Banana Is & How to Use It

Nano Banana is Google's name for a family of Gemini image models, not one permanent model ID. The current Google AI developer documentation identifies Nano Banana 2 as Gemini 3.1 Flash Image, Nano Banana 2 Lite as Gemini 3.1 Flash Lite Image, Nano Banana Pro as Gemini 3 Pro Image, and the earlier Nano Banana as Gemini 2.5 Flash Image. That distinction matters: the app may choose a model for you, while the API lets a developer choose an explicit model ID. Read the model name shown in the product you are using before making claims about cost, resolution or behaviour.

You can work in the Gemini app for conversational creation, Google AI Studiofor experiments and API keys, or the Gemini API and Vertex AI for applications, billing and governance. The same prompt may be accepted in all three surfaces, but the picker, quotas, image sizes and available features are not automatically identical. Do not copy a price from an old prompt library: Google publishes model-specific text and image pricing, and the model, resolution and billing surface determine the number. Free access is useful for learning, but a production workflow needs a quota check and an approval path.

The practical strengths are reference-led editing, multi-image composition, conversational refinement, product scenes, restoration and controlled visual variations. These are capabilities, not guarantees. Faces can drift, small labels can be wrong, hands and geometry can change, and a model can add plausible details that were not in the reference. Google says generated images include an invisible SynthID watermark. Keep the original asset, the prompt, the chosen model, and the final human review together. If a person is identifiable, obtain permission before uploading or publishing their likeness; if the image carries a claim, verify the claim outside the image model.

120+ Copy-Paste Nano Banana Prompts

Grouped by task. Start in the Gemini app or AI Studio, attach a reference where noted, and change one variable at a time so you can tell what improved the result.

Character-Consistent Edits

Nano Banana's signature move: change one thing while keeping the same person, pet, or object identical. Upload a reference photo first, then paste the prompt.

Keep this exact person identical, same face, hairstyle, and outfit, but change the background to a sunlit Tokyo street at golden hour. Match the lighting on their skin to the new scene.

Keep everything the same except the season: place this same person in the same pose but surround them with falling autumn leaves and warm October light.

Using this photo as the reference, generate the same character now wearing a tailored navy business suit instead of casual clothes. Preserve their exact facial features and expression.

Take this same product (the white ceramic mug) and show it on five different surfaces, marble counter, wooden desk, linen tablecloth, café table, and a windowsill, keeping the mug's shape, logo, and color identical in every shot.

Keep this same dog identical and create a three-panel sequence: sitting in a park, running on a beach, and sleeping by a fireplace. Maintain the exact fur pattern and collar across all three.

Multi-Image Blending & Merging

Feed Nano Banana two or more images and fuse them. Label your inputs (A = subject, B = background) so the model knows what to keep.

Image A is the main subject (a woman in a red coat). Image B is the background (a snowy mountain village). Place the subject from A naturally into scene B, matching the cool daylight, adding realistic snow on her shoulders, and grounding her with a soft shadow.

Blend these three product photos into a single clean flat-lay on a beige background, arranged as a balanced top-down composition with even studio lighting and subtle shadows.

Take the sofa from Image A and the living room from Image B. Replace the existing couch in the room with the sofa from A, matching perspective, scale, and the room's warm afternoon light.

Combine the face from Image A with the outfit and pose from Image B to create one cohesive portrait. Keep the identity from A and the styling from B, blended with consistent lighting.

Merge this landscape photo with this texture image: apply the watercolor paper texture from Image B as a subtle overlay across the landscape in Image A, keeping the scene fully recognizable.

Photo Restoration & Colorization

Repair and revive old or damaged photos. Upload the original, then describe the restoration you want.

Restore this old, faded photograph: remove the scratches, creases, and dust spots, sharpen the details, and rebuild any torn areas while keeping the original composition and the people's faces exactly as they are.

Colorize this black-and-white portrait with natural, period-accurate tones, realistic skin, hair, and clothing colors, without altering the facial features or expression.

Take this water-damaged photo and repair the stained corner, balance the exposure, and recover the washed-out highlights so the scene looks clean and evenly lit.

Enhance this low-resolution scan: increase the apparent sharpness and clarity, reduce grain and noise, and preserve every detail of the original people and setting.

Remove the date stamp and the unwanted person on the far left from this family photo, then fill the gaps naturally so the background looks seamless.

Style Transfer & Creative Restyling

Reimagine an existing image in a new visual style while keeping the underlying subject recognizable.

Restyle this photo as a hand-painted watercolor illustration with soft edges and visible brush texture, while keeping the subject, pose, and composition clearly recognizable.

Convert this portrait into a clean line-art ink drawing on white, preserving the exact facial features and hairstyle so it still looks like the same person.

Reimagine this street scene in a warm 1990s film-photography look: gentle grain, soft contrast, slightly faded colors, and a nostalgic glow, keeping the scene unchanged.

Transform this landscape into a detailed isometric miniature diorama, like a tiny crafted model, with soft studio lighting and a shallow depth of field.

Apply a clean flat-design vector illustration style to this product photo, simplifying it into bold shapes and a limited color palette suitable for a website hero.

Product Mockups & Marketing

Drop your product into polished commercial scenes for ads, listings, and social posts.

Place this product on a minimalist podium against a soft gradient backdrop with studio lighting and a gentle reflection, leaving clean negative space at the top for ad copy.

Create a lifestyle scene: show this skincare bottle on a marble bathroom shelf beside a folded white towel and a small green plant, in bright natural morning light.

Generate a flat-lay social media image of this product surrounded by complementary props on a pastel background, top-down, with even lighting and room for a caption.

Mock up this logo printed on a folded heather-grey t-shirt laid flat on a wooden table, with realistic fabric texture and soft daylight shadows.

Portraits, Couples, Families and Pets

Attach a reference image and define what must remain unchanged. Ask for one controlled change at a time; do not rely on a vague request for a perfect likeness.

Use the attached portrait as the identity reference. Keep the person's face, age, expression and natural skin texture unchanged. Replace only the background with a quiet, softly lit studio in warm grey. Crop as a professional headshot with shoulders visible.

Create a natural couples portrait from Images A and B. Preserve both people's real facial features and relative height. Place them walking beside a rainy city canal at blue hour, with believable reflections and no added text.

Use the family reference photo. Keep every person recognisable, preserve their age and clothing colors, and arrange them in a relaxed outdoor picnic scene. Do not add or remove people, pets or jewelry.

Turn this selfie into a clean professional profile photo. Keep my face and hair recognisable, replace the background with a pale neutral wall, use soft window light, and avoid changing my body shape or inventing accessories.

Create three vertical portrait variations from this same reference: candid café, shaded garden, and simple studio. Keep the subject's identity, hairstyle and eye color consistent across the set.

Use this child's photo only as a private family keepsake reference. Keep the child recognisable and fully clothed; create a gentle illustrated storybook scene with no identifying text, location or school branding.

Use the pet photo as the reference. Keep the animal's markings, muzzle shape, collar and size unchanged. Place it on a beach at sunrise with realistic paws, shadows and wind in the fur.

Create a passport-style crop from this portrait with a plain light background and even lighting. Preserve the person's actual features. Do not beautify, age, slim, widen or alter identity.

Products, Listings and E-commerce

Give the model product facts and an invariant list. For logos, labels and packaging, use the image as a concept and proofread against the real asset.

Use the attached product photo as the sole product reference. Keep its shape, color, logo placement and proportions exact. Create a clean 4:5 listing image on a warm white background with a soft shadow and generous negative space.

Place this exact bottle in a bright bathroom lifestyle scene. Keep the label blank if you cannot reproduce it accurately; do not invent claims, ingredients, measurements or certifications.

Generate five consistent product scenes from the same reference: marble, wood, linen, concrete and outdoor stone. Keep the product identical in every image and change only the surface and lighting.

Create a three-panel product comparison layout using the supplied objects. Preserve every object exactly, use equal scale, label panels A, B and C with plain text, and leave room for our verified copy later.

Show this item being used by an adult in a realistic home setting. Keep the product and its proportions unchanged. Do not imply a medical, safety, performance or financial outcome.

Turn this flat product photograph into a restrained editorial still life. Preserve the object and branding; add only a folded fabric, one plant and directional morning light.

Create a square social image for this product with a clear focal point and no text. Keep the object unchanged and leave the top third empty so a designer can add verified copy later.

Inspect this packaging mockup for likely text errors. Return the image concept and a separate checklist of every word that must be replaced with the approved artwork; do not claim the small text is accurate.

Infographics, Diagrams and Text

Treat image text as a draft. Supply short, exact labels, ask for a clean hierarchy, and verify every character before publishing.

Create a simple 16:9 process infographic with five numbered stages: Brief, Draft, Check, Approve, Publish. Use large high-contrast labels, one icon per stage, and no extra words.

Turn the supplied verified table into a clean editorial chart. Preserve every number and category exactly. If any label cannot be rendered confidently, leave an empty label box rather than guessing.

Create a classroom diagram of the water cycle for a 12-year-old audience. Use only these labels: evaporation, condensation, precipitation, collection. Keep arrows directional and avoid decorative text.

Design a restaurant menu board with exactly these three items and prices: [paste approved items]. Render the text large and plain; do not add ingredients, currency or promotional claims.

Create a technical architecture diagram from this list of components and connections. Use short labels, directional arrows, consistent spacing and a legend. Do not invent services or data flows not in the source.

Make a timeline from the supplied dates and events. Preserve the dates exactly in ISO format, use a readable horizontal layout, and leave any uncertain event unlabeled rather than filling the gap.

Create a clean map-style illustration of the supplied route without adding real-world place claims. Use only the provided place names and direction arrows, with a note area for later verification.

Create a bold thumbnail with the exact headline [paste headline], no other text, generous margins and strong contrast. After generation, return a separate transcription so I can compare it manually.

Scenes, Interiors and Consistent Series

Describe composition, camera position, materials and light. Keep the brief specific enough to steer the result while leaving room for the model to compose.

Create a cinematic 3:2 editorial photograph of a quiet Lagos street after rain, with warm shop lights reflected in the road, natural human proportions, restrained color and no readable brand names.

Illustrate a friendly robot helping a librarian organise books in a bright community library. Flat editorial shapes, warm paper texture, inclusive characters, no logos, no text.

Create a realistic interior redesign from this room photo. Keep the walls, windows and room dimensions unchanged. Replace only the rug, lighting fixture and seating with simple mid-century pieces.

Turn this landscape into a hand-painted travel poster. Preserve the major landforms and composition, simplify detail tastefully, and leave the title area blank for a designer.

Create a tasteful 1980s-inspired family living-room scene with period furniture, soft flash photography and believable materials. Do not use brand logos or copy a specific copyrighted poster.

Design a minimal app onboarding illustration showing a person choosing a calendar date. Use one consistent character, three flat colors, accessible contrast and no words except Next.

Create a close-up macro still life of the supplied flower. Keep its natural structure and color, use a dark neutral background and side light, and avoid adding insects, dew or petals that are not visible.

Create four matching seasonal versions of this composition: spring, summer, autumn and winter. Preserve camera angle, subject placement and scale; change only season, light and surrounding details.

8 Tips for Better Nano Banana Results

  • Start edit prompts with 'Keep everything the same except…' so Nano Banana preserves identity and only changes what you specify.
  • Use short, clear sentences. Concise instructions consistently beat long, complex paragraphs with this model.
  • When blending images, label them (A = subject, B = background) and say which one to keep, it removes ambiguity.
  • Edit conversationally: send one change, see the result, then refine in your next message instead of rewriting the whole prompt.
  • Name the lighting you want ('warm afternoon light', 'soft studio lighting') so merged or relocated subjects look believable.
  • For consistency across a series, reuse the same reference image and the phrase 'the same person/product' in every prompt.
  • Describe the surface and surroundings for product shots, material, color, props, to get commercial-grade mockups.
  • Remember every output carries an invisible SynthID watermark; use edits for enhancement and styling of images you own.

A reliable Nano Banana workflow

Start with a small brief: name the subject, the intended change, the composition, the light and the output shape. Then list the invariants. For a portrait, invariants might be identity, age, expression, hair and clothing. For a product, they might be the exact silhouette, label placement, color and dimensions. This is more useful than writing a paragraph of adjectives because it tells the model what is allowed to move.

Generate a first draft, inspect it at full size, and write down the first concrete failure. If the face changed, say that the face must remain unchanged. If the room geometry changed, lock the walls and windows. If the text is wrong, do not ask for more decorative detail; replace the text with a short approved string and verify it character by character. One correction per turn makes the cause of an improvement easier to understand.

For a repeatable series, save the original reference, the exact prompt, the model ID or picker label, aspect ratio, date and selected output. Reuse the same reference rather than feeding the previous generated image back into the next turn indefinitely. A generated image can accumulate small identity and geometry changes. The reference is your anchor; the output is a draft that still needs review.

For client, commerce or public work, add a human sign-off: check likeness consent, copyright or trademark exposure, product accuracy, readable text, accessibility and provenance. The image model can help with ideation and production drafts, but it is not a source of truth for prices, medical claims, locations, dates or legal permissions.

Nano Banana 2 or Pro?

Google describes Nano Banana 2, the Gemini 3.1 Flash Image model, as the general-purpose workhorse in the current API family. It is the sensible starting point for everyday generation, editing, multiple reference images, consistency and text. Nano Banana Pro, Gemini 3 Pro Image, is positioned for complex visual tasks, stronger world knowledge, precise creative control and higher-resolution output. Nano Banana 2 Lite is the efficiency option when throughput and latency matter more than maximum quality.

That is a selection rule, not a promise that one model will win every brief. Your choice should follow the acceptance test. If the task is a quick thumbnail with no important text, start with the general or Lite model. If the task contains a dense diagram, a product reference, complex spatial instructions or a print asset, compare against Pro at the resolution you actually need. Record the model and cost assumptions; model names and pricing can change.

The Gemini app may abstract these choices behind a product experience. The API exposes model IDs and pricing documentation. Never describe an image as “4K” merely because the prompt asked for 4K, and never describe a label as accurate without checking the rendered pixels against approved copy.

Primary sources and live details

Model names, availability, quotas and pricing are unstable. Google's documentation is the authority for the current API model IDs and image-generation behavior. Product surfaces can roll out at different times, so verify the picker and account you are actually using.

Nano Banana FAQ

What is Nano Banana?

Nano Banana is Google's nickname for its Gemini image generation family. The current API docs distinguish Nano Banana 2 (Gemini 3.1 Flash Image), Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image), Nano Banana Pro (Gemini 3 Pro Image), and the earlier Nano Banana (Gemini 2.5 Flash Image). The family supports generation, editing, multi-image input and conversational refinement.

Is Nano Banana free?

Access and quotas depend on the Google product, account, region and model. The Gemini app and AI Studio can offer free access, while API use is billed by model and output size. Check Google's live pricing page before budgeting a batch. Google says generated images include SynthID, so keep provenance and consent in your workflow.

How do I access Nano Banana?

Use the Gemini app for conversational creation, Google AI Studio for developer experiments, or the Gemini API/Vertex AI for an application and controlled billing. The exact model picker and limits differ by surface, so select the model where the product exposes it rather than assuming the app and API behave identically.

What is Nano Banana best at?

It is useful for controlled image generation and editing: reference-led portraits, product consistency, scene changes, multi-image composition, readable labels, restoration and iterative refinement. No prompt guarantees identity, factual text or exact geometry; inspect every output.

How does Nano Banana compare to other image models?

Do not treat a universal winner as a verified fact. Choose Nano Banana 2 for a general-purpose balance, Nano Banana Pro for complex visual work and higher resolution, and the Lite variant when speed and cost dominate. Compare current model documentation and your own acceptance criteria before committing a production workflow.

Can I edit my own photos with Nano Banana?

Yes. Upload a photo and describe the change in plain language, remove an object, swap the background, fix lighting, colorize, or restyle. Treat edits as enhancement and styling of images you own or have the right to use. Do not impersonate real people, create misleading content, or edit photos of others without consent. The built-in SynthID watermark helps keep AI-generated images identifiable.

Keep exploring

Don't stop here

What to read next

Hand-picked guides our readers explore right after this one.