nano-banana is the image generation and editing model for Google Gemini 2.5 Flash Image, designed to transform creative descriptions, product photos, and style references into visual assets. Its focus is on natural-language editing, multi-image blending, and subject consistency: you can create from text or progressively change scenes, adjust lighting, and replace objects around existing images, making it suitable for e-commerce and brand content that requires ongoing iteration.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
Creation and editing capabilities correspond to Gemini 2.5 Flash Image; aspect ratios, quantities, and result fields differ according to the selected platform endpoint.
Submit product photos together with style references, and explain the role of each image, such as preserving the product shape, borrowing background tones, or adjusting the scene layout. nano-banana can generate blended images around these cues, making it suitable for exploring new compositions based on existing brand assets rather than starting from a blank description every time.
When editing, you can directly describe “change the background to a wooden table,” “make the lighting soft morning light,” or “replace an object in the image.” A more practical approach is to also specify which elements should be preserved and focus each edit on a clear objective, making it easier to compare before-and-after differences while reducing the cost of repeatedly reworking the entire image.
When a set of assets needs to repeatedly feature the same product or character, continue using the same reference image and keep the appearance, color scheme, and style descriptions fixed. The model helps maintain subject recognizability, allowing scenes, lighting, and composition to become the primary variables, which is suitable for visual exploration of serial posters, brand character stories, and storyboard concepts.
Provide a clear product image, describe the target environment, camera angle, and lighting, and generate desktop, home, or outdoor scene assets. Then try different backgrounds and whitespace layouts around the same product to deliver a set of advertising candidate images for designers to review; product marks, packaging text, and key structures should be checked item by item before publication.
Using character reference images as a foundation, describe the actions, environment, and mood of each scene to create visual drafts for a continuous story. Writing character appearance requirements separately from scene changes allows the objective of each image to be controlled more clearly. The deliverables are suitable for discussing shots and narrative direction, rather than directly replacing precise character design artwork.
First prepare an accurate title and a small amount of supporting text, then specify the hierarchy, background, and whitespace placement to generate event posters, menu concepts, or social covers. Simple layouts make it easier to check the relationship between text and visuals; after selecting a composition, gradually adjust colors and element positions, and use design tools to finalize text when strict typography is required.
If tasks mainly involve product scene replacement, character variations, and creative drafts, nano-banana's fast generation and natural-language editing positioning are a good fit. It corresponds to Gemini 2.5 Flash Image, while nano-banana-2 corresponds to Gemini 3.1 Flash Image; they are not different spellings of the same model. When mature templates already exist, the base model can be used first to validate the visual direction.
nano-banana-pro corresponds to Gemini 3 Pro Image and is another version option. For refined brand visuals, use the same set of assets and prompts to compare candidate results, then decide which model to use based on text, structure, and detail requirements. Do not directly apply the resolution or tool capabilities of Pro or nano-banana-2 to nano-banana; parameter combinations should also be checked again when switching.
Choose generate for text-based creation; choose edit to modify existing assets, use image_urls to provide reference images, and separately specify what to preserve and what to change.
Specify model=nano-banana, action, and prompt for /nano-banana/images; start with aspect_ratio=1:1, resolution=1K, and count=1, setting aspect ratio and resolution separately.
Get the image from data[].image_url; for asynchronous requests, query with task_id or receive a callback. When continuing edits, pass the selected image again and narrow the scope of changes for each round.
Keep the white sneakers, laces, and side logo from the original image, changing only the background to a bright indoor wooden floor, with light entering from a window on the right, and do not add text.
Check the shoe shape, laces, and logo item by item; in the next round, correct only the background or lighting to avoid changing both the product and the scene at once.
No. nano-banana corresponds to Gemini 2.5 Flash Image; nano-banana-2 corresponds to Gemini 3.1 Flash Image. Nano Banana is the public nickname for the former. When integrating, use the full invocation ID to avoid mixing parameters and creative capabilities of different versions due to similar names.
When using /nano-banana/images, set action=generate and submit prompt for text-to-image generation; for editing, set action=edit, provide assets through image_urls, and describe the intended changes. It is recommended to explicitly specify model=nano-banana and describe the subject, scene, and elements that need to be preserved.
Multiple reference images can be used to guide blended creation. The key is to clearly state the purpose of each image, such as using the product image for its shape and the style image for its color palette and atmosphere. Choose assets with consistent goals whenever possible, specify the main subject in the prompt, and carefully check after generation whether product details have been affected by style changes.
You can use /openai/images/generations or /openai/images/edits, and explicitly set model=nano-banana. Submit prompt for generation; for editing, additionally submit the image URL or URL array for image. Asset fields and result fields differ between endpoints, so do not mix image_urls and image during integration.
Image URLs from the dedicated endpoint are located in data[].image_url and can be used for subsequent display or download; the OpenAI-style endpoint returns URLs or Base64 data through data. The API also provides task-processing parameters such as async and callback_url. Applications should distinguish image results from task_id to avoid treating successful task submission as completed image generation.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗
↗
↗This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs