🍌 Google AI · Image Prompts
Google Nano Banana
Prompts
Nano Banana is the funniest name in AI and one of the most useful tools in Gemini. It edits photos, makes new images and renders readable text, and the quality depends heavily on your prompt. This guide, part of our Google AI tools guide, explains the three models, gives you prompt templates that work, and shows how to edit instead of starting over.
Short answer: Describe the subject, action, setting and style, then edit by saying what changes and what stays the same. For text in images, write the exact words first. Use Nano Banana 2 for speed and Nano Banana Pro for text-heavy or complex work.
Worth knowing
- There are three models, so check which one your prompt is using.
- Google's own guide recommends writing the text you want first, then asking for the image.
- Editing prompts should say what changes and what stays.
The three models
| Model | Built on | Best for |
|---|---|---|
| Nano Banana | Gemini 2.5 Flash Image | The original, from 2025 |
| Nano Banana 2 | Gemini 3.1 Flash Image | Fast, everyday images, and the default in the Gemini app |
| Nano Banana Pro | Gemini 3 Pro Image | Text-heavy, complex or premium work |
Nano Banana Pro launched on November 20, 2025. Reports say Nano Banana 2 became the default image model in the Gemini app in February 2026. Names and defaults can change, so check the model picker.
A prompt formula for new images
Use this order: subject, action, setting, style, and detail. For example: “A small red kite flying over a harbor at sunrise, watercolor style, soft pastel colors, wide shot.”
Ten prompt templates
- “A [subject] in [setting], in the style of [medium], with [mood] lighting.”
- “Using this photo, change the sky to a sunset and keep everything else the same.”
- “Create a poster that says ‘Open Mic Night’ in bold block letters on a dark blue background.”
- “Make this product photo look like it's on a marble counter, with soft morning light.”
- “Turn this sketch into a clean line drawing.”
- “Show the same character from the side, from behind and from the front.”
- “Create an infographic that compares three ways to learn a language.”
- “Make this room look like it was decorated in a minimalist style.”
- “Restyle this photo as a children's book illustration.”
- “Combine these two reference images: put the product from the first into the room in the second.”
For playful experiments without any image tools, see our complete list of Google Easter eggs.
Editing, not restarting
When you have a base image, focus on what's changing and what's staying. “Change the jacket to green, keep the face, pose and background” works better than describing the whole image again. Make one change at a time and build up.
Text in images
- Write the exact words. Put them in quotation marks.
- Draft the text first. Chat with Gemini to refine the wording, then ask for the image.
- Name the style. “Bold block letters” or “handwritten script.”
- Keep it short. Fewer words render more reliably.
Consistency and safety
Google's guide says Nano Banana models can mix up to 14 reference images to keep a character or product consistent. Generated images carry a SynthID watermark and C2PA Content Credentials. Don't use a real person's likeness without their permission, and follow Google's usage rules.
If you're choosing between image tools, our compare AI tools page can help. Other tools such as Midjourney and Claude's text-based prompts for scene ideas can also feed your workflow.
How we checked this: We used Google Cloud's ultimate Nano Banana prompting guide from March 2026, Leonardo.ai's prompt guide, Pixelbin's model overview and Google Cloud's product pages. Model names and defaults changed several times in 2026. AI products change quickly, so check dates and your own screen before you rely on any detail. If something here is out of date for you, email ownerflexora.ai@gmail.com and we will fix it.
Your turn: Take one of your own photos and ask for one precise edit. Then ask for a second edit, and notice how the specific wording changes the result.
Good image prompts read like a photographer's brief, and the best habit is to change one thing at a time.
More tricks to try
Working with documents instead of pictures? The Google NotebookLM tips guide is next.