Multi-reference product scene
Example prompt
Use reference 1 for the exact product, reference 2 for the marble set, and reference 3 for warm sunset lighting; centered commercial composition, no extra products.
credits : 1
Nano Banana 2 Image Generator
Create an image from a prompt or use multiple references to guide the subject, product, composition, wardrobe, color, or lighting in one image-to-image task.
How to Use Nano Banana 2
Use text alone or assign a clear purpose to each reference image before generating.
- Choose text-to-image, or upload 1–8 reference images for image-to-image.
- Write a prompt that names the role of each reference, then choose Auto or one of the fixed aspect ratios.
- Generate and inspect identity, hands, text, logos, and details that were supposed to remain unchanged.
Nano Banana 2 Image Generator FAQs
What is Nano Banana 2 used for?
Use it for text-to-image generation or reference-guided transformations such as product composites, wardrobe studies, character variations, social layouts, and campaign concepts.
How many reference images can Nano Banana 2 use here?
Image-to-image accepts from one to eight reference images. Assign a clear role to each image, such as subject, product, composition, lighting, or style, to reduce conflicting guidance.
How do I get better Nano Banana 2 results?
Describe the subject, composition, lighting, style, camera angle, and constraints. For multi-reference work, explicitly state what to take from each image and what must not change.
What prompt and aspect-ratio limits are available?
The prompt supports up to 5,000 characters. You can use Auto plus ten fixed ratios, including square, portrait, landscape, vertical, and ultrawide formats.
Does image-to-image change only the part I describe?
Not necessarily. This is prompt-guided generation rather than a mask-based layer editor, so nearby details can also change. Check the full image after every transformation.
What should I review before publishing a result?
Check faces, hands, small text, logos, product details, reference-image rights, and any brand-specific requirements. Multiple references can also introduce unwanted extra subjects.
Current settings and limits
Text generation and multi-reference image transformation with up to eight inputs.
- Input
- Text or 1–8 reference images
- Workflows
- Text to image and image to image
- Prompt limit
- Up to 5,000 characters
- Aspect ratios
- Auto plus 10 fixed ratios
Model result Gallery
No public result is stored for this Hub yet. Use the examples below to create the first model-specific result.
Model-specific prompt examples
Character wardrobe study
Example prompt
Keep the character identity and face from the first image, apply the jacket from the second image, and use the studio lighting from the third image.
Campaign ratio variation
Example prompt
Create a clean summer beverage campaign visual with room for a headline, citrus accents, bright hard shadows, and a composition that remains readable in vertical format.
Know before you generate
Reference conflicts
Using many images can introduce competing subjects or styles; assign a clear role to each reference.
Not a layer editor
Prompt-guided changes can affect nearby details, so inspect elements that were meant to stay unchanged.
Production review required
Check faces, hands, logos, small text, brand details, and usage rights before publishing.