Start with the destination ratio
Choose the publishing surface and format before writing the full prompt. Compose for the final crop from the beginning instead of repairing a generic square image afterward.
15CREDITS
Remaining credits：
Public
Create and Edit Images with Nano Banana Pro
Nano Banana Pro combines text-to-image creation and reference-guided editing in one model workspace. Begin with a prompt or add up to eight input images, select a square, portrait, landscape, or automatic aspect ratio, choose 1K, 2K, or 4K resolution, and export PNG or JPG. Keeping these controls together makes it easier to test a concept, preserve important source details, and prepare a file for its actual destination.
Nano Banana Pro image creation controls
Build each request from a clear creative goal, relevant source images, the destination format, and the resolution required for review or delivery.
Prompt-based image generation
Create a new image from a written brief covering subject, setting, composition, lighting, style, materials, and constraints. Start with the most important visual decision and keep secondary details organized.
Up to eight reference images
Use references for people, products, locations, layout, palette, or texture. Explain the role of each upload and remove conflicting sources so the model receives one coherent direction.
Automatic and fixed aspect ratios
Let the model choose automatically or select a common square, portrait, or landscape format. Set the destination before generation to avoid losing essential subjects or text during a later crop.
1K to 4K, PNG or JPG
Match resolution and file type to the workflow. Use faster review files while exploring, then choose a larger PNG or JPG once the composition and required details are approved.
How to prepare a clear image request
Strong inputs make the desired change and the protected details easy to distinguish. Plan the final placement, reference roles, and review checks before rendering.
1
Define the outcome
Write down the audience, channel, ratio, resolution, file type, and single most important message. This prevents decorative details from overtaking the purpose of the image.
2
Describe the visual hierarchy
Name the primary subject, supporting elements, background, viewpoint, lighting, palette, and empty space. State which element should attract attention first and where any text must remain readable.
3
Organize reference priorities
Assign each uploaded image to identity, product geometry, pose, layout, color, or surface detail. If sources conflict, state which reference controls the final decision.
4
Compare controlled revisions
Review composition, identity, object shape, edges, text, and crop safety separately. Change one instruction at a time and keep the strongest result as the next editing base.
A repeatable creation and editing workflow
01
02
Generate or edit with clear priorities
Use text mode for a new concept and reference mode when existing visual details must guide the result. Separate required changes from protected elements in the prompt.
03
Finish at the right resolution and format
After the design is stable, select 1K, 2K, or 4K and choose PNG or JPG according to the delivery need. Inspect small details and text at the final output size.
Input requirements and output options
Check the current upload and export limits before building a request. Matching the source files to these controls reduces avoidable failures and keeps the prompt focused on the creative decision rather than file preparation.
01
Up to eight input images
Reference mode accepts a maximum of eight uploaded images. Select only the sources needed to explain identity, product geometry, scene structure, pose, color, or texture. Fewer purposeful references are easier to prioritize than several near-duplicates. For a new concept without protected source details, choose text mode and begin without an upload.
02
30 MB maximum per image
Each source image can be up to 30 MB. Confirm file size before submission when using high-resolution photography or design exports. If compression is needed, preserve the features the model must read, such as facial structure, product edges, labels, material texture, and the boundaries of any element that should remain unchanged.
03
1K, 2K, and 4K resolution
Choose from 1K, 2K, or 4K in the generator. A smaller output is practical for comparing concepts and prompt changes; a larger output is more appropriate after composition and required details have been approved. Resolution changes output dimensions, but it cannot resolve ambiguous instructions or incompatible references on its own.
04
PNG and JPG delivery
The current workspace offers PNG and JPG output. Choose PNG when the next step benefits from a broadly supported lossless working file, and JPG when a smaller delivery file suits the publishing workflow. Always inspect the final image at its intended display size, because format choice does not correct text, edges, identity, or crop issues.
Select a mode for the result you need
Text generation and reference-guided editing solve different production problems. Start from whether the result should be invented from a brief or anchored to existing visual material.
01
Create from text when the direction is open
Use text-to-image for a fresh scene, illustration, campaign direction, or concept frame. Define the subject, setting, visual hierarchy, viewpoint, lighting, palette, and destination ratio. Keeping the prompt organized makes it easier to compare alternative interpretations and identify which instruction should change in the next attempt.
02
Edit from images when continuity matters
Use image-to-image mode when identity, product structure, pose, room layout, or another source detail must guide the output. Assign each reference a role, state what may change, and list the protected elements separately. This gives the request a clearer hierarchy when several images contribute different information.
03
Choose settings from the destination
Decide the publishing surface before generation, then select the aspect ratio, resolution, and PNG or JPG output that fit it. A marketplace image, vertical social post, presentation graphic, and print draft require different framing. Designing for the final crop early is more reliable than removing important content afterward.
04
Use the displayed credit amount
Kuca shows the current credit requirement inside the generator for the selected model mode and settings. That dynamic value is the reliable reference because costs and available configurations may change. Recheck it whenever you switch mode, resolution, or another option, rather than relying on a price copied into static page text.
Practical uses for reference-guided creation
Ecommerce imagery
Place a product in controlled scenes, test background and lighting directions, and keep product shape, materials, and brand details explicit.
Social creative
Compose vertical, square, and landscape visuals around one campaign idea while preserving safe space for platform text and interface overlays.
Portrait and character edits
Use reference images to guide identity, wardrobe, pose, or setting, and state clearly which facial and character details must remain stable.
Brand exploration
Compare palettes, materials, environments, and art directions without losing the hierarchy, product geometry, or message of the original brief.
Concept art
Develop environments, props, characters, and key frames from a structured description, then refine selected directions with targeted references.
Print and high-resolution delivery
Move an approved visual to 2K or 4K and select PNG or JPG based on the final editing, review, and distribution workflow.
Frequently asked questions about modes and outputs
Nano Banana Pro supports prompt-based image generation and reference-guided editing. You can add up to eight images, choose an aspect ratio, render at 1K, 2K, or 4K, and select PNG or JPG output.
The model accepts up to eight reference images. A smaller, well-labeled set often gives clearer direction, especially when different sources control identity, product details, composition, and style.
Yes. Select the text variant to create from a prompt, or the image variant to upload references and direct an edit. Both variants are available from the model page selector.
You can choose 1K, 2K, or 4K. Use a smaller output while testing ideas, then increase resolution after the composition and required details are stable.
Yes. Nano Banana Pro offers PNG and JPG output. Choose the format according to whether the next step favors editing flexibility, file size, or direct publishing.
Describe what must change and what must stay fixed, give each reference one role, remove contradictory sources, and revise one variable at a time. This makes each result easier to compare and diagnose.
Start your next image request
Generate from text or guide an edit with reference images, then choose the format, resolution, and file type you need.