Nano Banana Pro prompt guide
The fastest way to consistent results is to see how others do it. Sign in to browse works and prompts shared by other users in the gallery, then copy and adapt them to your needs. Below are the basics of writing a good prompt.
Capabilities
Core capabilities
Built around real creative workflows — get going with plain natural language.
Subject
State what to generate or keep — product, person, room, character or sketch.
Scene
State background, space, time, lighting and composition.
Style
Specify photography, illustration, comic, photoreal render or brand visual.
Constraints
Spell out what must be kept, must not change and should be avoided.
Purpose
Tell the model if it is for a hero image, poster, avatar, detail page or proposal.
Keep asking
If the first take is off, add "change only X" and "keep Y" by chatting.
Workflow
Workflow
Write a one-line goal
e.g. "turn this product shot into a premium e-commerce hero".
Add what to keep
e.g. "keep the product shape, logo, color and material".
Specify the purpose
e.g. "for a D2C homepage banner, leave space on the right for copy".
Use Cases
Use cases
E-commerce hero template
Keep the product, switch to a premium studio background, soft shadows, for detail pages.
Interior render template
Keep the layout, switch to modern wood style, add daylight and light furnishings.
Character consistency template
Keep face, hair color and outfit palette; generate different poses and expressions.
Style transfer template
Keep the composition, convert to soft Japanese illustration or photoreal commercial photography.
Sketch coloring template
Complete material, color, light and background from the line art.
Local editing template
Change only the chosen area; keep the rest of the subject and background.
In depth
A good prompt isn't a pile of fancy words — it's stating subject, scene, style, camera, lighting and text clearly. Nano Banana Pro understands natural language well, so describe things the way you'd talk; the key is complete information in a clear order, so the model knows what's in the frame and what matters most.
Practical structure: state the subject and action first, then environment and style, then modifiers like camera, lighting and aspect ratio; put copy in quotes when you want text; add a reference image to keep a character or product consistent. When editing, describe only the part to change rather than rewriting everything. Multiple languages are supported.
The examples below span posters, comics, architecture and concept art — read them to feel "how a one-line request maps to an image," then adapt to your own scenarios.
Showcase
Real examples
Real work from JiaoDesign users, covering the scenarios this page is about.

Poster copy
Quote the headline; key visual and text generated together.

Text comic
Prompt style for panels plus dialogue, lightweight narrative.

Sequential art
Fixed character plus continuous story prompts.

Concept mood
Mood and lighting words to control the image's emotion.

Comic layout
Example combining panels and art-style description.

Arch viz
Breaking down structure, material and lighting prompts.
FAQ
FAQ
Does natural language work well?
Are longer prompts better?
How do I stop the model changing the subject?
What if the first take is off?
Start creating with JiaoDesign now
Free credits on signup, shared across models, pay as you go.