Nano Banana Prompts: A Copy-Paste Library for Gemini Image Edits and Designs

Nano Banana prompts are prompts written for Nano Banana, Google's name for the image models in Gemini. In the Gemini app, Nano Banana 2 makes the picture whenever the chat is set to Flash or Pro. The 18 prompts below follow the shapes in Google's own prompting guide: edits to your own photo, pictures built from two or three photos, cards and labels with words, film and camera looks, and light and color fixes. The two photo sections at the top and the image-quality section at the end need a photo of your own. The blocks also run in Toybox AI's Image Generator, though the same prompt gives a different result in each tool.

How a good Nano Banana prompt is built

A good Nano Banana prompt reads like a short scene description, not a tag list. Google's Nano Banana prompting guide describes each job like this:

• New image: subject, action, location, composition, then style. For example: "A retired fisherman mending a green net on a harbor wall at sunrise, close-up from a low angle, documentary photo style."
• Image from your photos: the photos, what each one contributes, then the new scene.
• Edit: the change, then a list of what must stay exactly the same.

The guide also asks you to open with a strong verb, to describe what you want rather than what you don't ("an empty street", not "no cars"), to use camera words such as "low angle" or "aerial view", and to put any words for the image in quotes. Fill in this template to edit a photo of your own:

Using my photo of [subject], [change it how]. Place it in [new setting] with [light]. Keep [features that must not change] exactly the same. Style: [photo, painting, illustration].

For the model's versions and what each one is for, read what Nano Banana is.

Gemini image prompts for your own photo

These Gemini image prompts start from one photo you upload. In the Gemini app, attach the photo to the chat. Google allows photo edits there only for users 18 and older. In Image Generator, the photo goes in through "Add reference photos". Every prompt ends with a keep-list, because Google's guide asks edits to spell out what stays the same. Toybox's privacy policy says your prompts, uploads and results aren't used to train AI models.

Same house, snowy evening

Shows your home in another season, useful for holiday cards.

Using my photo of the front of our house, show the same house on a snowy winter evening: snow on the roof and the lawn, warm light in every window, a pine wreath on the front door. Keep the house's shape, colors, windows and front steps exactly the same.

Change: the season, such as spring with cherry blossoms along the walk.

1990s school portrait

The retro portrait trend, with the studio details that make it read as the decade.

Using my photo, make a 1990s school picture day portrait: a blue laser-beam backdrop, soft studio flash, a slight film grain and a shoulders-up pose turned a little to one side. Keep my face, hairstyle and glasses the same.

Change: the decade, such as a 1970s studio portrait with a mottled brown backdrop.

More versions of this trend, decade by decade, are in the AI yearbook photo prompts.

Dog as a formal oil portrait

A painted pet portrait that keeps the markings owners check first.

Using my photo of my dog, paint him as a formal oil portrait in a dark green velvet jacket with gold buttons, seated in a carved wooden chair against a dark background, lit like a museum painting. Keep his fur color, markings and ear shape the same.

Change: the costume, such as a captain's coat and a ship's wheel behind him.

Living room restyle

A redesign that keeps the room's bones so you can judge the furniture.

Using my photo of this living room, restyle it with a warm mid-century look: a walnut sideboard, a mustard armchair, a round wool rug and a globe floor lamp. Keep the walls, windows, floor and camera angle exactly the same.

Change: the style, such as a coastal look with white linen and pale wood.

Prompts that combine photos or keep a character

The first two prompts merge two photos, and the last two carry one toy or one drawing into a new picture. Google's launch post says Nano Banana 2 can keep as many as five characters and 14 objects consistent in one workflow, while its developer docs give the API limits for the same model as four characters and 10 objects. Image Generator takes up to three photos through "Add reference photos"; add them first photo first. Only use photos of other people with their permission.

Cat moved onto a new sofa

Places a pet from one photo into the room from another.

Place the gray tabby cat from my first photo on the green velvet sofa in my second photo, curled up on the middle cushion. Match the room's warm afternoon light and add a soft shadow under the cat. Keep the cat's stripes, eye color and size the same.

Change: the pet and the spot, such as a dog on a porch swing.

Two people on one bench

Puts two people photographed separately into one candid shot.

Put the two people from my photos side by side on a park bench in autumn, both smiling at the camera, with yellow leaves on the path and soft late-afternoon light. Candid photo style. Keep each person's face, hair and glasses as they are in their own photo.

Change: the place, such as a beach boardwalk at sunset.

Plush toy on an adventure

Keeps one original toy consistent in a new scene, the first step toward a storybook series.

Using my photo of this plush rabbit as the character, show it sitting in a tiny wooden canoe on a calm pond at sunrise, with mist on the water and reeds along the bank. Keep its floppy ears, button eyes and blue knitted scarf exactly the same.

Change: the adventure, such as the rabbit flying a paper airplane over a garden.

Google's own Nano Banana tips post singles out turning photos into figurines as an especially popular prompt. That desk-figure trend has its own 3D figurine prompts.

Drawing to a real treehouse

Turns a sketch into a realistic photo while keeping its layout.

Use my drawing of a treehouse as the layout and turn it into a realistic photo: the same ladder, windows and roof shape, built from weathered cedar planks in a large oak tree, late afternoon sun through the leaves.

Change: the build, such as a backyard shed or a playhouse.

Prompts for cards, labels and signs with words

Google says Nano Banana 2 renders accurate, legible text for greeting cards and marketing mockups, and can translate text inside an image. Put every word in quotes and keep each line short.

Thank-you card for a coach

A card with two lines of text in different sizes.

A thank-you card for a kids' soccer coach with the words "Thanks, Coach Delacroix" in bold rounded letters across the top and "From the Brambleskip Tadpoles" in smaller script below. An illustrated soccer ball and whistle, bright green and white colors, lots of white space.

Change: the team, the sport and the colors.

Recipe card with four steps

A small infographic with short numbered lines, which keep each step easy to proofread.

An illustrated recipe card titled "Weeknight Tomato Soup" with four numbered steps in short lines: "1. Soften the onion", "2. Add tomatoes", "3. Simmer 20 minutes", "4. Blend and season". A small drawing beside each step, cream background, clean sans-serif text.

Change: the recipe and the drawings.

Jam jar label

A product label with a name, a size line and a border.

A product label for a jar of homemade peach jam that reads "Aunt Delphine's Peach Jam" with "Small batch, 8 oz" underneath. Hand-drawn peaches around the edge, a vintage letterpress look, cream and coral colors.

Change: the flavor, the name and the colors.

Bilingual café chalkboard

Two languages on one sign, the job Google says Nano Banana 2 can handle with in-image translation.

A café chalkboard sign that reads "Soup of the day: Lentil" on the top line and "Sopa del día: Lentejas" on the bottom line, in white chalk lettering with small drawn curls of steam, a wooden frame around the board.

Change: the dish and the second language. Check every accent in the result.

Gemini image prompts with an aesthetic look

These aesthetic Gemini image prompts lean on a camera or film look. Google's guide says naming the hardware changes the result: a disposable camera brings the harsh, nostalgic look of on-camera flash, and asking for slightly grainy 1980s color film gives a nostalgic, gritty feel.

Rainy window seat on film

A soft, faded film look with one warm light.

A dreamy 35mm film photo of a young woman in a cream knit sweater reading on a window seat on a rainy afternoon. Soft gray daylight from the window, a warm lamp glowing in the corner, gentle film grain and faded colors.

Change: the activity, such as writing in a journal with a cup of tea.

Disposable camera party flash

A harsh-flash snapshot, the look of photos from a paper camera.

A disposable camera flash photo of friends laughing around a backyard picnic table at night, string lights in the background, paper plates and a half-eaten cake on the table. Harsh direct flash, red-tinged shadows, slightly tilted frame, 1990s color film look.

Change: the gathering, such as a campsite or a kitchen dance party.

Noodle shop on a wet night

A moody night scene seen from across the street.

A moody night photo of a tiny noodle shop seen from across a wet street: steam fogging the front window, a cook ladling broth under one bare bulb, a red paper lantern by the door, and its glow reflected in the puddles. Teal-tinted shadows and warm highlights, like cinematic color grading.

Change: the place, such as a laundromat or a corner bookshop.

Nano Banana prompts to enhance image quality

These prompts ask the model to redraw your photo with better light, color or less noise. A redraw is not a rescan: the model invents detail it cannot see, so compare faces and any small text with the original before you share the result. For a photo that is only blurry, Image Unblur in Toybox does that one job, and it also rebuilds detail rather than recovering it. For scratches, tears and faded prints, the old photo restoration prompts go further.

Faded print restored

Brings back color and removes damage on a scanned family photo.

Using my scan of this faded 1970s family photo, restore it: bring back natural color, remove the dust spots and the crease across the top corner, and sharpen the faces slightly. Keep every person, pose, outfit and the background the same.

Change: the fix, such as only removing a coffee stain.

Window-light relight

Replaces harsh indoor light on a portrait with soft daylight.

Using my photo, relight it as if it was taken beside a large window on a bright overcast day: soft even light on the face, no harsh shadows, natural skin texture. Keep my face, expression, hair and clothes the same.

Change: the light, such as warm golden hour from behind.

Night photo cleanup

Cuts the grain in a dark phone photo without losing the evening mood.

Using my photo taken at night, reduce the grain and noise and bring out detail in the shadows, while keeping the warm glow of the streetlights and the dark sky. Keep the people, the buildings and the colors the same.

Change: the scene, such as a candlelit birthday dinner indoors.

Habits that improve Nano Banana results

These five habits come from Google's developer docs and its Nano Banana prompting guide.

• Keep editing in the same chat. Google's developer docs call multi-turn conversation the recommended way to iterate: ask for "the same, but warmer light" instead of rewriting the prompt.
• Settle the words first. Google's docs say text comes out best when Gemini writes the words in one turn and draws the picture in the next. Ask for three headline options, pick one, then request the card.
• Name the lens and the light. The guide pairs a low camera angle with an f/1.8 aperture for shallow focus, and suggests a three-point softbox setup for even product light.
• Name the material and the color. Google's guide asks for "navy blue tweed" rather than just "suit jacket", and the same habit applies to a sofa, a mug or a scarf.
• Mind the download size. Google's help page puts Gemini app downloads at 1K on an account with no Google AI plan and 2K on a paid plan, so a free download suits a phone screen more than a print.

Fixes for common problems

Google's docs say Gemini matches the shape of the photo you give it, or makes a square when there is no photo, unless you ask for another shape. For more edits, see the photo editing prompts. For new pictures from text alone, browse the AI image prompts library, and for other tools that do this job, compare the Nano Banana alternatives.

Try these prompts in Image Generator

Each block has a "Try this prompt" link that loads it into Image Generator, ready to run. For the photo prompts, first add up to three JPG or PNG photos, 10 MB each, with "Add reference photos": the one the prompt calls "my first photo" goes in first. Image Generator follows your photos loosely rather than copying them, so compare faces, markings and lettering with the originals before you share the result. An image currently costs 50 credits. For cheaper 1K drafts in 14 shapes, including Social 4:5, use Image Generator Lite, currently 35 credits, and to put a new scene behind a person in a photo, use Background Swapper.