Use Nano Banana Pro for complex visual briefs that demand high fidelity, clear typography, real-world context and careful reference handling. Generate at 1K, 2K or 4K from text or up to five uploaded images.
Write a precise design brief, select the destination ratio and resolution, then add references when identity, product shape, styling or composition must remain consistent.
Nano Banana Pro is selected automatically. The model switcher remains available for side-by-side prompt comparisons.
Google positions Nano Banana Pro, also known as Gemini 3 Pro Image, as the premium model in its image family for complex composition, precise text and professional creative control.
Nano Banana Pro was introduced as Google's studio-quality image generation and editing model built on Gemini 3 Pro. Its official demonstrations concentrate on context-rich infographics, legible multilingual text, consistent multi-reference composition and directed changes to focus, lighting, camera angle and color. Those are practical strengths when an image must communicate a system or preserve several supplied ingredients, not simply look attractive.
The ReelCine integration keeps the workflow direct: write a prompt, optionally upload up to five images, choose one of seven common aspect ratios and generate at 1K, 2K or 4K. Google's full API can accept larger reference sets and supports additional tool-assisted workflows, but those broader limits should not be confused with the five-image upload control offered on this page.
The official examples below make each claim visible. The plant card turns a photograph into a structured explainer. The astronaut storyboard holds one narrative together across several shots. Localized cans show translated packaging in context. Fourteen distinct characters sharing a room demonstrate how the model can reconcile a crowded visual brief without collapsing everything into a generic scene.
The generator on this page exposes the practical controls below. They are narrower than Google's complete API surface and describe what you can use here today.
Use references to carry people, products, logos, layouts or materials into a new composition.
All three tiers are available without a separate quality selector.
Create square, portrait, landscape and vertical assets in the format required by the destination.
1K and 2K cost 4 credits; 4K costs 5 credits per image.
The examples are official Google outputs selected to match the specific behavior described beside them.
Turn a supplied subject and an information brief into a visual explainer with headings, labeled modules and supporting illustrations. The String of Turtles example uses the plant photograph as evidence, then organizes origin, care and growth information around it.
Create with Nano Banana Pro
Plan several camera views while maintaining the same character, setting and narrative intent. The Urban Astronaut sequence includes an establishing shot, medium shot, close-up and point-of-view frame in one coherent board.
Create with Nano Banana Pro
Translate more than isolated labels by adapting readable copy to a product or campaign composition. The paired beverage cans demonstrate English and Korean packaging while retaining the visual identity of the pack.
Create with Nano Banana Pro
Bring many recognizable subjects into one setting while keeping their distinctive colors, forms and personalities. Google's living-room example arranges fourteen fluffy characters around a television without losing the ensemble.
Create with Nano Banana ProA good reference edit identifies the unchanged geometry before requesting a new atmosphere. The official fox example changes daytime to night while holding the subject, camera and snowy forest stable.
Turn this scene into nighttime. Keep the fox in the same position, preserve the snowy birch forest and camera angle, and replace the warm daylight with believable cool moonlight and deep blue shadows.
The model is best justified when fidelity, information, language and reference consistency all matter to the final asset.
Combine people, wardrobe, products and setting references into one editorial direction. State which identities and objects must remain recognizable, then define the camera, lighting and final format.
Create recipe cards, process diagrams and explainers that depend on both visual sequence and readable labels. Verify every instruction and fact before distributing the image.
Integrate a word into architecture, packaging or a physical scene instead of placing plain copy over a background. Describe the material, spatial relationship and intended readability.
Separate required source material from the new creative direction. The model responds better when every uploaded image has a named role and the intended output has a clear use.
Deliverable and audience + required subjects and reference roles + composition and camera + exact copy and language + lighting, materials and grade + ratio and constraints + details that must remain consistent.
Create a 16:9 premium outdoor campaign image using the uploaded model, jacket and trail-shoe references. Preserve the model's identity, the exact jacket color blocking and the shoe silhouette. Place her on a windswept black-sand ridge at blue hour, walking toward camera from a low three-quarter angle. Natural fabric movement, cool ambient light with one warm horizon glow, room on the left for a headline, no added logo and no text.
Say which image defines identity, wardrobe, product, layout or style. An unlabeled stack of references forces the model to guess how their evidence should be combined.
Name the camera angle, focal point, depth of field, lighting direction and color grade. These concrete choices are more useful than asking for a cinematic result without explanation.
Quote exact headlines and translations, explain their hierarchy, and keep the first layout proofreadable. Correct important copy in a focused second pass.
Begin with a clear production goal, use references only when they add essential evidence, and review the 4K result as closely as any commissioned asset.
Define the asset, audience, message, composition and finish. Include all required copy and distinguish hard constraints from creative preferences.
Upload up to five useful references, select the destination ratio, and choose 1K or 2K for drafts or 4K for detailed final work.
Review every face, product feature, label, hand and small object. Refine one category at a time so successful parts of the composition are not needlessly disturbed.
The current ReelCine integration uses resolution-led pricing and does not add a separate quality selector.
| Control | Available value | When to use it |
|---|---|---|
| Resolution | 1K, 2K, 4K | Use 1K for quick direction, 2K for most digital placements and 4K when small detail or larger output matters. |
| Aspect ratio | 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9 | Match the delivery frame before generation to avoid cropping away carefully arranged people, copy or products. |
| Reference input | Optional, up to 5 images | Upload only references that carry necessary identity, object, layout or style information. |
| Credits | 4 at 1K/2K, 5 at 4K | The current cost appears before submission, and ReelCine refunds credits when a generation fails. |
Pro is the quality-led Google option. Use a lighter model when the job is exploratory, repetitive or primarily constrained by speed and volume.
| Decision | Nano Banana Pro | Consider another model |
|---|---|---|
| Complex multi-reference scenes | A strong fit when several people, products or visual ingredients must remain recognizable in one composition. | Choose Nano Banana 2 for faster high-volume variations once the visual system is already established. |
| Infographics and localization | Built for knowledge-rich graphics, long-form visual text and multilingual adaptation. | Compare Seedream 5.0 Pro when information architecture and directed design transformation are the main challenge. |
| Studio-style control | Well suited to detailed instructions about camera, focus, lighting, color and composition. | Use GPT Image 2 when transparent output or explicit Standard, Medium and High quality tiers are more useful. |
| Iteration cost | The current 1K and 2K options cost 4 credits, with 4K at 5 credits. | Nano Banana 2 starts at 2 credits and is better suited to larger batches of exploratory directions. |
High fidelity reduces cleanup, but it does not remove the need for factual, legal and visual quality control.
Nano Banana Pro is Google's premium image generation and editing model, also known as Gemini 3 Pro Image. It is designed for high-fidelity, knowledge-rich and professionally controlled visual creation.
Yes. Upload a source image and describe what should change as well as the identity, geometry, camera, product or layout details that must remain fixed.
Yes. The ReelCine generator offers 1K, 2K and 4K. The 4K tier costs 5 credits, while 1K and 2K currently cost 4 credits.
This interface accepts up to five images. Google's underlying model documentation describes broader reference capabilities, but the five-image product limit is what applies here.
It is specifically positioned for legible text, localization, diagrams and context-rich infographics. Important facts, spelling and translations still need human verification.
The model is designed for high-fidelity multi-reference composition. Results improve when each reference is clear and the prompt explicitly states which identities and attributes must be preserved.
No. Google's own products and API documentation describe grounded workflows, but this page currently exposes prompt and reference-image generation rather than a Search-grounding control.
A 1K or 2K image costs 4 credits and a 4K image costs 5 credits. The current total is shown before generation and failed jobs are refunded.
Use Nano Banana 2 when speed, volume and low iteration cost matter more than the Pro model's maximum fidelity and control. It is especially practical for testing several directions before a premium final pass.
Model facts and official examples come from Google DeepMind's launch material and current Gemini documentation. The page describes ReelCine's own controls separately where they differ from Google's wider API surface.
Give every reference a purpose, choose the final frame early, and let Nano Banana Pro resolve the full brief into one polished image.
Create with Nano Banana Pro