Nano Banana prompt generator for images and photo edits
This Nano Banana prompt generator writes clear, sentence-based instructions for Google's Gemini image model, both for creating new pictures and for editing a photo you upload. Edit mode adds a preservation clause so the model changes only what you ask and keeps the person, pose and framing as they were.
Your optimized prompt
🔒 Runs in your browser. Your text is not stored on our servers.
How to use it
Create or edit
Leave the edit box unticked to create a new image, or tick “Edit an existing photo” if you will upload a picture to Gemini.
Describe the image or the change
For a new image, describe the subject. For an edit, write the single change you want, such as a new background or outfit color.
Add style and details
Choose style, lighting, camera and mood, and add text to render, things to avoid and the output aspect ratio.
Paste into Gemini
Copy the prompt and use it in the Gemini app, Google AI Studio or any tool that offers the Nano Banana model; attach your photo for edits.
Why Nano Banana prefers sentences
Nano Banana is the popular name for Google's Gemini image generation and editing model. Because it is built on a language model, it understands descriptions that read like a brief: who is in the scene, where, how it is lit and how it should feel. Keyword piles that work in older diffusion tools tend to get flatter results here. The builder therefore writes full sentences in a fixed order: what to create, how the shot is framed, the light, the mood, the palette, your details, any text and the aspect ratio.
Create a cinematic film still of a street food vendor flipping noodles in a busy night market. The scene is captured as a wide-angle shot, with strong leading lines. It is lit by neon lighting. The overall feeling is joyful. Use a magenta and teal color palette. Steam rising from the wok, motion in the crowd. Avoid text, logos. Aspect ratio 16:9.
Edit mode: change one thing, protect the rest
Most disappointing photo edits happen because the model “improves” things you did not mention: a slightly different face, a new pose, a changed crop. When edit mode is on, the builder starts with “Using the provided image,” followed by your instruction, and then adds an explicit preservation sentence:
Using the provided image, change the background to a quiet beach at sunset. Keep everything else identical — the person's face and identity, pose, body proportions, clothing, framing and the original lighting — unless I mention it. Match the original photographic style and perspective. Do not add: text, extra people. Output aspect ratio 4:5.
If you pick a style, the prompt asks to match that look instead of the original one; leave style empty for the most faithful edit. Lighting and details you add are appended as extra instructions. Edits work best when each prompt asks for one change. For a sequence (new background, then new outfit, then color grade), generate a separate prompt for each step and apply them one after another in the same conversation.
Good edit instructions versus vague ones
| Vague | Specific |
|---|---|
| make it better | brighten the face slightly and remove the red cast from the skin |
| change the clothes | replace the grey hoodie with a navy wool blazer over a white shirt |
| put her somewhere nice | place her in a sunlit café with blurred tables in the background |
| add a sign | add a wooden sign above the door that reads "OPEN" |
Name the object, the change and, when it matters, the material or color. For words on the image, use the text field: the builder asks the model to render them clearly and spelled exactly.
Combining photos and other workflows
Nano Banana can also work with more than one reference image, for example a product photo plus a scene. To build that kind of prompt from your own pictures, use two images to prompt, which writes a “Using the two provided images…” instruction. If you want a structured version for repeatable edits, tick the JSON option here or try image to JSON prompt. For the same idea in other models, the image prompt generator switches format with one click.
Please edit only photos you have the right to use, and do not use edits to impersonate or mislead. Google states that images made with its models carry an invisible SynthID watermark.
Pro tips
- For edits, describe only the change; the preservation clause already covers everything else.
- Leave the style menu empty in edit mode when you want the result to look like the same photo.
- Ask for one edit per prompt and chain them; multi-change prompts drift more.
- Put exact words for signs or labels in the text field and keep them short.
- If the face still changes, repeat “keep the face identical” in your own instruction and try again.
Frequently asked questions
What is Nano Banana?
It is the widely used nickname for Google's Gemini image generation and editing model, available in the Gemini app, Google AI Studio and other Google tools. Model names and versions change, so check Google's documentation for the current one.
Does this tool upload my photo?
No. The Nano Banana prompt generator only writes text. You attach the photo yourself in Gemini, so the image never passes through this site.
How do I keep the same face when editing?
Use edit mode. The prompt explicitly tells the model to keep the face and identity, pose, proportions, clothing, framing and lighting identical unless you mention them. One change per prompt also helps.
Can Nano Banana write text on images?
It handles short text well. Add the words in the text field and the prompt will ask for them to be rendered clearly and spelled exactly. Check spelling in the result and regenerate if needed.
Is this site affiliated with Google?
No. It is an independent free tool. Gemini and Nano Banana are Google products.
Related tools
Updated: