Nano Banana Prompts: Copy, Edit, and Create Images

Nano Banana is Google’s name for its Gemini image-generation models. The current Gemini image-generation documentation describes a workflow that accepts text, images, video, or a combination of inputs, then supports conversational editing. The product details in this article were checked on 2026-09-20. That makes prompt quality less about finding one magical sentence and more about giving the model a clear brief, inspecting the image, and asking for a focused change.

This page collects Nano Banana prompts you can copy and adapt. They are reusable examples for planning and iteration; they are not presented as results generated on this site. Model names, limits, and availability can change, so check Google’s current Gemini image-generation documentation before integrating an API workflow.

A prompt structure that stays editable

A useful prompt usually answers five questions:

What is the subject? Name the person, product, place, or scene.
What should the viewer see? Describe pose, camera angle, framing, and important details.
What visual treatment fits? Specify medium, lighting, color, texture, or art direction.
Where will the image be used? State aspect ratio, negative space, or a layout need.
What must remain unchanged? When editing a reference image, identify the elements to preserve.

For example:

Create a clean product photograph of a matte black travel mug on a pale stone table.
Use soft morning window light, a three-quarter camera angle, realistic materials,
and generous empty space on the left for a headline. Keep the mug logo simple and
legible, with no extra objects in the background. Use a 4:5 portrait composition.

The prompt describes the product, scene, lighting, composition, layout, and constraint in one pass. If the result is close but the background is too busy, the next request can change only that issue.

Copy-paste prompts for common tasks

Text-to-image concept

Create a cinematic editorial illustration of a lone cyclist crossing a foggy bridge
at dawn. Use a muted teal and amber palette, clear depth between foreground and
background, realistic wet pavement, and a calm reflective mood. Compose it as a
16:9 hero image with open space on the right for a short title.

Product mockup

Create a realistic studio mockup of this packaging design on a white pedestal.
Use soft directional light, subtle contact shadows, accurate paper texture, and a
three-quarter view. Keep the supplied brand mark and all label text unchanged.
Do not invent extra logos, badges, or claims. Use a 1:1 square composition.

Edit a supplied image

Edit the supplied photo by replacing the cloudy sky with a warm late-afternoon sky.
Keep the person, clothing, building edges, camera perspective, and color of the
foreground unchanged. Match the new light direction to the existing shadows and
blend the horizon naturally.

Add or remove an object

In the supplied kitchen photo, remove the blue plastic bowl from the counter.
Reconstruct the counter surface and keep every other object, reflection, and camera
angle unchanged. Preserve the original warm indoor lighting.

Diagram or storyboard

Create a four-panel storyboard for a short product demo. Panel 1 shows a close-up
of the device, panel 2 shows a hand pressing one button, panel 3 shows the result
on a laptop, and panel 4 shows the product on a clean desk. Keep the same character,
props, color palette, and camera direction in every panel. Leave a small caption area
under each frame and use clear, readable placeholder text only.

Start using Nano Banana 2

Multi-turn editing that preserves intent

Gemini’s image workflow supports conversational changes. Treat the first generation as a base and make each follow-up request narrow. “Keep the subject, camera angle, and palette; change the jacket from red to olive” is easier to evaluate than a new paragraph that restates the entire scene with different adjectives.

A practical sequence is:

Generate the composition with the subject, setting, and framing.
Correct the largest visual issue, such as pose, layout, or lighting.
Add a small styling detail or remove a distracting element.
Request final output size and aspect ratio after the content is stable.

When a detail must remain accurate, repeat the preservation instruction in the edit request. For a reference image, identify which objects are authoritative and which parts may change. If the model alters a logo or a person’s identity, provide a clearer reference and ask it to preserve those features explicitly.

Reference images, text, and layout constraints

Google’s current documentation describes support for multiple reference images and image editing. Use a short role for each reference: “product shape,” “material texture,” “character face,” or “color palette.” This prevents the model from guessing which image should control which detail. If you need text inside the image, write the exact wording, capitalization, and placement. Then inspect every character at the final output size; generated typography can still need a separate design pass.

For marketing layouts, state the destination: social card, portrait poster, app banner, or presentation slide. Ask for negative space on a named side and specify the approximate crop. A prompt that explains the layout is more useful than adding generic quality adjectives.

Choosing a model and output size

Google’s documentation currently lists Nano Banana 2 (Gemini 3.1 Flash Image), Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image), Nano Banana Pro (Gemini 3 Pro Image), and the legacy Nano Banana model. The right choice depends on the task, speed, cost, and detail requirements. Use the current model documentation for exact limits and API names rather than copying a stale configuration from a blog post.

The same documentation describes aspect-ratio controls and output sizes up to 4K for Gemini 3 image models. Choose the aspect ratio that matches the publishing surface, then request a larger size only when the image will be cropped or printed. A larger output cannot repair an unclear composition, so settle the subject and layout first.

A final review checklist

Before saving a Nano Banana result, check the subject, hands and faces, logos and labels, text spelling, edges around edited objects, and the crop at the actual display size. Compare the result with the prompt’s preservation requirements. Keep the prompt beside the image so the next edit can be precise instead of starting over.