Nano Banana 2.1 Prompt Guide
21 Nano Banana 2.1 prompts to copy or run in one click — headshots, photo editing, product shots, posters with text and multi-image edits — plus what’s new in 2.1 and how to fix a prompt that isn’t working.
- Prompt length
- Up to 8,000 characters on Moluri
- Modes
- Text to image, or attach up to 14 reference images to edit and combine
- Output
- 1K, 2K or 4K, in 14 aspect ratios or Auto, up to 4 images per run
- Languages
- Google lists English and 14 other languages — including Spanish, Japanese, Hindi and Chinese — for best results
- Cost to try
- 2 / 3 / 7 credits per image at 1K / 2K / 4K (Ext edition: 2 / 2 / 3)
- Iterate cheaply
- Test ideas at 1K, then run the keeper at 2K or 4K — and use 2K or above whenever the image contains text
What’s new in Nano Banana 2.1 for prompting
Nano Banana 2.1 is Google’s update to Nano Banana 2, released on October 6, 2026. You write prompts the same way; what changes is how much you can ask for.
- Text you can rely on
- Google built 2.1 for clear text on posters and detailed diagrams, including text in several languages. Quote the exact words and use 2K or 4K.
- Characters that stay the same
- Google rates 2.1 for keeping up to 4 characters and 10 objects consistent. In Google’s own ratings it scored 1106 on multi-character consistency, ahead of Nano Banana Pro (1011) and Nano Banana 2 (978).
- Finer control over edits
- Google designed 2.1 for editing with fine control over several quick rounds. Describe the one thing to change and say what must stay.
- Your old prompts still work
- Google documents one set of prompting guidelines for all Nano Banana models. Prompts written for Nano Banana 2 or the Gemini app carry over unchanged.
It isn’t flawless: Google’s model card lists blurry small text (especially at 1K), occasional left–right mix-ups, poses that stick from the reference image and imperfect likeness — fixes for each are in Prompt not working? below.
The Nano Banana 2.1 prompt formula
Write full sentences, not tag lists, and cover these six parts in roughly this order. A quick sketch can be one sentence; a finished image deserves all six.
- Subject
- Who or what is in the picture, and what they’re doing. Put it first.
- Setting
- Where it happens and what’s around the subject.
- Light
- Time of day and the quality of the light — soft window light, hard noon sun, neon at night.
- Camera
- Shot type, angle and lens: a close-up at eye level on an 85mm lens, an overhead flat lay, a wide shot from below.
- Style
- Photo, illustration, 3D render or painting — and the mood or color treatment.
- Purpose
- What the image is for: a product listing, a website header, a poster. The model uses this to fill in what you didn’t spell out.
A barista pouring latte art into a white ceramic cup at the counter of a small corner café, early-morning sun coming in low through the front window and catching the steam. Close-up from just above the counter on a 50mm lens, shallow depth of field. Warm, natural colors with a light film grain. For the hero image on the café’s website, so leave some empty space on the right.
Aspect ratio and resolution are settings in the generator, so you don’t need to write them into the prompt.
21 prompts, grouped by what you want to make. Copy one, or press “Try this prompt” to load it and its settings into the AI image generator; for the ones marked with a note, add your own photo first.
Portrait and headshot prompts
Start from a clear, front-facing photo and say that the face must not change — then describe the outfit, background, light and framing you want.
Professional headshot prompt
Turn the person in this photo into a professional headshot for a LinkedIn profile. Keep their face, hairstyle and expression exactly as they are. Navy blazer over a plain white shirt, soft studio lighting, plain light-gray background, framed from the chest up, looking at the camera. Natural skin texture, no heavy retouching.
Add your own photo as a reference image first.
- 1:1
- 2K
Instagram profile picture prompt
Turn the person in this photo into a bright, friendly Instagram profile picture. Keep their face, hairstyle and skin tone exactly as they are. Head-and-shoulders crop centered in a square frame, a relaxed genuine smile, soft natural light and a warm pastel backdrop in peach and coral. Light film grain.
Add your own photo as a reference image first.
- 1:1
- 2K
Realistic street portrait prompt
Candid street photo of an elderly man selling oranges from a wooden cart at a market in Seville, laughing with a customer just out of frame. Mid-morning sun, shot at eye level on a 35mm lens, natural colors, light film grain, realistic skin texture.
- 3:4
- 2K
Photo editing prompts
Attach the photo, name exactly what should change, and end with what must stay the same. One change per run gives the cleanest results.
Enhance image quality prompt
Recreate this photo as a sharp, high-resolution version. Remove the blur, noise and compression artifacts and bring back fine detail in faces, hair and textures. Keep the composition, colors, people and every object exactly the same — don’t add or remove anything.
Add your own photo as a reference image first.
- Auto
- 4K
Restore an old photo prompt
Restore this old family photograph. Remove the scratches, creases, dust and stains, repair the torn corner and bring back the faded areas. Keep every person’s face, expression, pose and clothing exactly as they are, and keep the original framing. Then colorize it with natural, period-appropriate colors.
Add a scan or photo of the old print as a reference image first.
- Auto
- 2K
Change the background prompt
Replace the background with a quiet beach at sunset. Match the light, shadows and color of the person to the new scene so it looks like one photo. Keep the person, pose, clothing and framing unchanged.
Add your own photo as a reference image first.
- Auto
- 2K
Remove unwanted objects prompt
Remove the people and the parked cars in the background so the street is empty. Fill the gaps with matching cobblestones and shopfronts. Keep the main subject, the light and everything else exactly as it is.
Add your own photo as a reference image first.
- Auto
- 2K
Swap one object prompt
Change only the white coffee mug on the desk to a matte green ceramic mug of the same size, in the same position. Keep the desk, the laptop, the light and everything else exactly as it is.
Add your own photo as a reference image first.
- Auto
- 2K
Change the style prompt
Turn this photo into a hand-painted watercolor illustration with loose brushstrokes, soft washes of color and visible paper texture. Keep the composition and keep the main subjects in the same places.
Add your own photo as a reference image first.
- Auto
- 2K
Product photography prompts
Describe the surface, props and light like a photographer would. When you start from your own product photo, protect the label and logo by name.
Skincare product photo prompt
A high-end product photo of an amber glass serum bottle with a gold dropper, standing on a wet slate stone among a few eucalyptus leaves and fine water droplets. Soft light from above with a gentle rim light, shallow depth of field, sharp focus on the bottle. Clean, luxurious look for an online store listing.
- 4:5
- 2K
Product photo from your own shot prompt
Place the product from this photo on a sunlit marble kitchen counter beside a bowl of lemons and a folded linen towel. Soft morning window light, shallow depth of field. Keep the product’s shape, colors, label and logo exactly as they are. Clean e-commerce look.
Add a photo of your product as a reference image first.
- 4:5
- 2K
Several products in one scene prompt
Arrange the three products from the reference images on a pale stone pedestal against a soft sage-green backdrop — the tallest one at the back, the two smaller ones in front. Even studio light with gentle shadows. Keep each product’s shape, color and label exactly as in its reference.
Add one reference image per product.
- 1:1
- 2K
Poster, sign and infographic prompts
Clear text is one of Nano Banana 2.1’s strengths. Quote the exact words, say where they go, and use 2K or 4K so small lettering stays sharp.
Travel poster with text prompt
A vintage travel poster for Lisbon with the word “LISBON” in large, bold, cream-colored serif letters across the top and “Summer 2027” in smaller letters underneath. Below, a yellow tram climbs a steep cobblestone street at sunset. Flat screen-print texture, limited palette of orange, teal and cream.
- 2:3
- 2K
Bilingual café sign prompt
A hand-painted chalkboard sign outside a small café. At the top, “TODAY’S SPECIAL” in white capital letters; below it, “Café de olla · Mexican spiced coffee” in yellow script; “$4.50” in the bottom right corner. Chalk texture, a few doodled coffee beans, warm afternoon light on a brick wall.
- 3:4
- 2K
Simple infographic prompt
A clean, flat infographic titled “HOW COLD BREW IS MADE” with four numbered steps from left to right: 1 “Grind coarse”, 2 “Add cold water”, 3 “Steep 12–24 hours”, 4 “Strain and serve”. One simple icon per step, a cream background, dark brown text and generous spacing.
- 16:9
- 2K
Multi-image prompts
Attach up to 14 reference images and refer to them by order — “the first image”, “the second image” — and by what they show.
Put a product on a model prompt
Put the sneakers from the first image on the person in the second image. Full-body shot on a city sidewalk in soft overcast light, with shadows and perspective that match the scene. Keep the person’s face and the sneakers’ design, colors and logo exactly as they are in the references.
Add two reference images in this order: the product, then the person.
- 3:4
- 2K
Two people in one frame prompt
Seat the person from the first image and the person from the second image side by side on a park bench in autumn, laughing at something off camera. Medium shot, warm late-afternoon light, a few falling leaves. Keep both faces, hairstyles and outfits exactly as in their references.
Add one photo of each person, in the order they’re mentioned.
- 3:2
- 2K
Same character, new scene prompt
Show the character from this image riding a bicycle through a rainy, neon-lit street at night, seen from a low side angle. Keep the character’s face, hair, outfit and art style exactly the same.
Add one image of your character first.
- 16:9
- 2K
Trending and fun prompts
The formats people share most — figures, expression grids and stickers. Use a photo of yourself, a friend who agreed to it, or your pet.
Collectible figure prompt
Turn the person in this photo into a 1/7-scale collectible figure standing on a round clear acrylic base on a wooden desk. Behind it, a computer screen shows the figure’s 3D model, and beside it sits a printed toy box with the figure’s artwork. Keep the face, hairstyle and outfit recognizable. Realistic product photo, soft indoor light.
Add your own photo as a reference image first.
- 4:5
- 2K
3×3 expression grid prompt
Create a 3×3 grid of nine portraits of the person in this photo, each with a different expression: laughing, surprised, thoughtful, winking, serious, shy smile, annoyed, excited and sleepy. Same lighting, same plain light-gray background and same outfit in every panel. Keep the face identical in all nine.
Add your own photo as a reference image first.
- 1:1
- 2K
Cute sticker prompt
A die-cut sticker of a sleepy sloth hugging a takeaway coffee cup. Cute cartoon style with bold, clean outlines, flat pastel colors and a thick white border, on a plain white background.
- 1:1
- 1K
Nano Banana 2.1 prompt templates
Copy a template, replace the brackets, and delete anything you don’t need.
Photo-real scene
A [shot type] photo of [subject doing something] in [setting]. [Light and time of day]. Shot on a [lens] from [angle]. [Mood or color treatment]. For [what the image is for].
Product shot
A [style] product photo of [product — material, color, finish] on [surface or backdrop], with [a few props]. [Lighting setup] for [the effect you want]. [Camera angle] to show [the feature that matters]. Sharp focus on [key detail].
Portrait from your photo
Turn the person in this photo into [type of portrait] for [where it will be used]. Keep their face, hairstyle and expression exactly as they are. [Outfit], [background], [lighting], framed [crop].
Poster, sign or infographic
A [poster / sign / infographic] for [brand or topic] with the text “[exact words]” in [font description], placed [where]. [Imagery]. [Style and colors].
Edit one thing
Change only the [element, described precisely] in this image to [new element]. Match the light, shadows and perspective of the photo. Keep everything else exactly as it is, including [the parts that matter most].
Combine several images
Take the [element] from the first image and [put it on / place it in] the [element] from the second image. [The final scene: setting, light, angle]. Keep [details that must not change] exactly as they are in the references.
Tips for editing, reference images and text
Editing a photo
Attach a photo as a reference image and the generator switches from creating to editing. The prompt only needs to cover the change.
- Name the change precisely
- “Change the blue velvet armchair by the window” beats “change the chair”. If there are several of something, describe the one you mean.
- Say what must stay
- End with what to keep: “Keep the person, pose, clothing and framing unchanged.” Without it, the model may tidy up things you liked.
- Say how it should blend
- Ask for the new element to match the light, shadows and perspective of the photo, so it looks like it was always there.
- Refine one step at a time
- On Moluri each generation is separate. To keep improving an image, download it, attach it as the reference for the next run and describe only the next change.
Combining reference images
- Refer to images by order and content — “the woman in the first image”, “the jacket from the second image” — and upload them in the order you mention them.
- Give each image one job: one for the person, one for the product, one for the setting.
- Protect what matters. For a face or a logo, say it must stay exactly as it is and describe its key features.
- Set the aspect ratio. With Auto, an edit follows the shape of your reference; when you combine images of different shapes, pick the ratio you want.
Getting text right
- Quote the exact words and say where they go: the headline “OPEN LATE” across the top in bold red letters.
- Describe the font rather than naming one: “a bold condensed sans-serif”, “elegant hand-lettered script”.
- Write the words yourself. Decide the wording first, instead of asking the model to invent a slogan and render it in the same step.
- Keep it short and use 2K or 4K — headlines and labels, not paragraphs.
Nano Banana 2.1 prompt not working?
The most common problems and how to fix them — including the limitations Google lists in the Nano Banana 2.1 model card.
- It ignored part of the prompt
- Tag lists like “girl, city, night, 8k, best quality” give the model little to work with. Rewrite them as plain sentences, with the most important detail first.
- The edit changed too much
- Everything you don’t protect is up for grabs. Finish edit prompts with what must stay unchanged, and change one thing per run.
- The face doesn’t look like me
- Use a sharp, well-lit, front-facing photo and say “keep the face exactly as it is”. Google notes that likeness isn’t always perfect — run it again or try a different photo.
- The text is blurry or misspelled
- Long paragraphs and small text are known weak spots, especially at 1K. Keep the wording short, put it in quotes, use 2K or 4K, and check the spelling.
- Left and right got swapped
- Google notes that the model sometimes mixes up left and right. Add a second clue: “the mug on the left, next to the laptop”.
- The pose won’t change
- When editing, the model occasionally keeps the pose from your reference. Describe the new pose clearly, or add a second reference image that shows it.
- Too many people drifted
- Google rates 2.1 for up to 4 consistent characters per image. With more, faces are more likely to drift — check them and regenerate.
- It didn’t know about recent events
- On Moluri, Nano Banana 2.1 doesn’t search the web, so “today’s weather in Paris” won’t reflect today. Write the facts you need into the prompt.
- The request was refused
- Google applies safety filters, and some requests are blocked. On Moluri, rejected jobs are refunded automatically.
- Same prompt, same miss
- If a prompt misses twice, change it instead of re-running it. Adjust one thing — the light, the angle, a detail — so you can see what made the difference.
Nano Banana 2.1 prompt FAQ
What prompt should I use for Nano Banana 2.1?
There’s no single best prompt. The ones that work describe the subject, setting, light, camera and style in full sentences and say what the image is for. Start from one of the 21 examples above and change the details to fit your idea.
How do I prompt Nano Banana 2.1 correctly?
Describe the picture the way you’d brief a photographer: subject first, then setting, light, camera angle and lens, style, and purpose. When editing a photo, say exactly what to change and what to keep unchanged.
Do Nano Banana 2 prompts work in Nano Banana 2.1?
Yes. Nano Banana 2.1 is Google’s direct update to Nano Banana 2, and Google documents one set of prompting guidelines for all its Nano Banana models. Your prompts carry over unchanged — and those that rely on text or several characters may come out better, since Google says 2.1 improves both.
What are good Nano Banana prompts to enhance image quality?
Attach the photo and ask for a sharp, high-resolution version with the blur, noise and compression artifacts removed — and say that the composition, people and objects must stay exactly the same. Choose 4K. It works by recreating the image rather than upscaling the original pixels, so check faces and small details closely.
Why is my Nano Banana prompt not working?
The usual causes are a tag-list prompt instead of sentences, an edit that doesn’t say what to keep, long or small text at 1K, and relying on left and right alone. See “Prompt not working?” above for fixes to each.
What’s the ideal prompt for a profile picture?
Upload a clear, well-lit, front-facing photo. Ask to keep your face, hairstyle and expression unchanged, then describe the outfit, background, light and crop — for example “head-and-shoulders crop, soft natural light, warm pastel backdrop”. For LinkedIn, ask for a blazer, a plain gray background and studio light instead.
Are Gemini prompts the same as Nano Banana prompts?
Yes. Nano Banana is the name for the image generation built into Google’s Gemini models, so an image prompt written for the Gemini app works here too. On Moluri each generation is separate rather than a chat: to refine a result, attach it as a reference image and describe only the next change.
Is Nano Banana censored?
Yes, Google applies safety filters to Nano Banana, so some requests are refused. Moluri’s Acceptable Use Policy applies as well, and rejected jobs are refunded automatically.
How many reference images can I use?
Up to 14 on Moluri, each up to 20 MB. Google rates Nano Banana 2.1 for keeping up to 4 characters consistent and up to 10 objects faithful in one image. Refer to the images by order and by what they show.
Where can I use Nano Banana 2.1?
Google offers it in its own products, including the Gemini app, Google AI Studio and the Gemini API. Google hasn’t released the model for download, so it can’t be run locally. Moluri is an independent web app that also runs it, with no setup or API key needed; Moluri isn’t affiliated with Google.
How much does it cost to try these prompts?
On Moluri, an image costs 2 / 3 / 7 credits at 1K / 2K / 4K, or 2 / 2 / 3 with the Ext edition. Failed jobs are refunded automatically. Plans start at $15 a month for 400 credits — about 133 images at 2K.
How this guide was made
Based on the prompting guidance in Google’s Gemini API image generation documentation and the ratings and known limitations in the Nano Banana 2.1 model card from Google DeepMind, checked against the limits and settings of Moluri’s own Nano Banana 2.1 integration. The example prompts were written for this guide. Last reviewed October 16, 2026.