Guides · Generative AI

Gemini Nano Banana Photo Editing Prompts: 12 Copy-Paste Prompts and a Simple Formula

By Ravindra Bagale, trainer · LinkedIn · Published: 19 October 2026 · Last updated: 19 October 2026 · 11 min read

Nano Banana is the name of Google's image model inside the Gemini app. To use it, open Gemini, choose Create images from the tools menu, upload a photo, and type what to keep, what to change, and the style, light and size you want. The free plan uses Nano Banana 2 (Nano Banana 2 Lite when Gemini is set to Flash-Lite). A Google AI plan unlocks Nano Banana 2.1, Google's newest image model (October 2026). Every image gets a visible watermark and an invisible SynthID mark.

1. Suppose we run a small cake shop in Pune

Suppose we are Sneha. We run a home cake business in Pune and sell on Instagram and Swiggy. We have a simple phone photo of a chocolate cake on our kitchen table. The background is messy: a gas stove, a steel plate and a phone charger.

We want a clean Diwali post. We open Gemini and type a weak prompt first:

make this photo better

What happens: Gemini guesses. It may change the cake's colour, add random flowers, or even change the shape of the cake. The cake in the new picture is not our cake any more.

Now we type a strong prompt:

Keep the cake exactly the same: same shape, same chocolate colour, same decoration.
Change only the background: replace the kitchen with a soft cream wall and a few
small lit diyas on the left side, slightly out of focus.
Style: clean product photo, warm evening light from the right.
Size: square 1:1 for Instagram.
Add the text "Happy Diwali" in gold at the top. Do not add any other text.

Expected result (yours will be similar, not the same): the same cake on a clean cream background, with warm diya light and "Happy Diwali" at the top. Nothing about the cake has changed.

The second prompt worked because it told Gemini what must stay the same. That is the biggest secret of photo editing with AI.

2. What is Nano Banana, in simple words?

Nano Banana is the nickname of Google's Gemini image model. It can do two jobs:

1. Create a new image from your words (text-to-image).
2. Edit a photo you upload, using normal sentences (image editing).

Old photo apps needed you to select areas and move sliders. Here you just say, "Change the background to a beach". The model understands the photo and the sentence together.

There are a few versions. As of October 2026, Google's own Gemini page says:

Plans, names and limits change often. Check the Gemini app's own page before you pay for anything.

3. How to open Nano Banana in Gemini (step by step)

1. Open the Gemini app on your phone, or go to gemini.google.com on a computer.
2. Sign in with your personal Google account. You must be 18 or older for these image features.
3. Tap the tools button near the message box.
4. Choose 🍌 Create images.
5. Tap + to upload a photo, or skip this step to create an image from text only.
6. Type your prompt and send it.
7. If you do not like the result, do not start again. Type a follow-up in the same chat, like "Keep everything, but make the light warmer".

4. The 5-part formula for photo prompts

Every good photo prompt answers these 5 questions:

You can copy this blank template and fill it in:

Keep: [what must stay exactly the same]
Change: [what to add, remove or replace]
Style: [the look you want]
Light and camera: [light, angle, distance]
Size: [1:1 / 9:16 / 16:9]
Do not: [anything you do NOT want, like extra text or extra people]

The Do not line is optional, but it saves a lot of retries.

5. Worked example: one product photo, three edits

Let us go back to Sneha's cake photo and make three posts from the same photo, in the same chat.

Edit 1, a clean Instagram post: we used the prompt from section 1.

Edit 2, a Swiggy menu photo. We type a follow-up:

Keep the cake exactly the same. Now make a version with a plain pure white
background and no text, like a food delivery app menu photo. Size 4:3.

Edit 3, a WhatsApp status. One more follow-up:

Keep the cake the same. Make a vertical 9:16 version with the cream wall and diyas.
Leave empty space at the bottom for text. Add "Order on 98XXXXXX10" in white,
small, at the bottom centre.

Expected result (yours will be similar): three pictures of the same cake, each in the right size for the place we post it.

Check before you post: zoom in on any text. AI image models still make spelling mistakes in small text. Google's own model card for Nano Banana 2.1 lists "blurry small text" as a known limit. If the text is wrong, ask again or add the text yourself in Canva or your phone's editor.

6. 12 copy-paste photo prompts (with the formula marked)

Upload your photo first, then paste one of these. Change the words in [brackets].

Portraits

1. Retro saree portrait: "Keep my face and expression exactly the same. Change my outfit to a [red] chiffon saree. Style: 1990s Bollywood film still, soft grain. Light: warm golden light from the window. Size 4:5."
2. Professional LinkedIn photo: "Keep my face, hair and skin tone exactly the same. Change my clothes to a navy blazer and white shirt, and the background to a plain light grey office wall. Style: realistic corporate headshot. Light: soft front light. Size 1:1. Do not smooth my skin."
3. Festival look: "Keep my face the same. Add a traditional Maharashtrian nath and a green Paithani saree. Background: a decorated home entrance with a rangoli. Light: evening diya light. Size 9:16."
4. Black and white classic: "Keep everything in the photo the same. Turn it into a high-contrast black and white portrait, like an old film camera photo. Size: same as the original."

Fun and trends

5. Figurine on a desk: "Turn the person in this photo into a small realistic collectible figurine standing on a wooden office desk, next to a laptop. Keep the face recognisable. Style: product photo of a toy. Size 1:1."
6. Cartoon version: "Keep the pose and clothes colours the same. Turn this photo into a 3D cartoon movie style character. Background: same place, but cartoon style. Size 1:1."
7. Pet as a superhero: "Keep my dog's face, fur colour and markings exactly the same. Add a small red superhero cape. Background: a city rooftop at sunset. Style: fun movie poster. Size 4:5."

Fixing real photos

8. Remove a person in the background: "Keep me and everything else the same. Remove the man in the blue shirt standing behind me on the right, and fill that area naturally."
9. Fix lighting: "Keep the photo the same. It is too dark. Make it look like it was taken in bright daylight, with natural colours. Do not change any faces."
10. Restore an old family photo: "This is an old scanned family photo with scratches. Remove the scratches and dust, sharpen it gently, and keep every face exactly as it is. Keep it black and white."

Business

11. Shop banner: "Create a wide 16:9 banner for a kirana store called '[Patil Kirana Stores]'. Show fresh vegetables, rice bags and a smiling shopkeeper. Text at the top: '[Diwali Offer: 10% off]'. Style: bright, clean poster. Do not add any other text."
12. Combine two photos: upload two photos, then: "Put the person from the first photo standing next to the scooter from the second photo, outside a Pune street shop. Keep the face and the scooter exactly the same. Light: afternoon sun. Size 4:5."

7. Why does my face change, and how do I fix it?

The model draws a new image every time. It does not cut and paste your old pixels. So small details like eye shape, a mole, or a hairline can drift.

8. Use it safely: watermarks, privacy and the law

• Every Gemini image has a visible watermark and an invisible SynthID watermark. Do not try to remove them. You can upload an image to the Gemini app and ask if it was made with Google AI. Gemini checks the SynthID mark.
• Only upload photos of people who agreed. Editing a friend's or a stranger's photo without permission can hurt them, and fake images of real people can be a crime.
• India's rules on AI images. The IT Rules amendment of 2026 (in force from 20 February 2026) says platforms must clearly label AI-made ("synthetically generated") images and must stop deceptive deepfakes. Treat your AI photos as AI photos. Say so when you post them if they look real.
• Never use an AI-edited photo for official documents like Aadhaar, PAN, passport, a visa or a job ID. Those need a real, unedited photo.
• Do not upload ID cards, bank details or children's school photos to any AI tool.

9. Beginner mistakes and quick fixes

10. Try at home

Recap

Pick one real photo today, run it through the 5-part formula, and post the result. That one try will teach you more than ten reels.

Learn the basics first: What Is Generative AI and How Does ChatGPT Work? and Prompt Engineering for Beginners: Role, Task, Context and Format

Ravindra Bagale, trainer: linkedin.com/in/ravindra-bagale. The people, shops, phone numbers and offers on this page are made up for learning. Swiggy is named only to make the situation real. Gemini plans, model names and limits were checked on Google's Gemini pages on 19 October 2026 and change often. AI images change on every try, so your result will be similar, not the same.