nano-banana-2:official is the image generation and editing entry point for Google Nano Banana 2, designed for creative tasks that require repeated adjustments to composition, text, and product scenes. It corresponds to Gemini 3.1 Flash Image, enabling both image generation from text descriptions and modification of existing content using reference images. It is suitable for event posters, e-commerce assets, and series visual production, keeping creation focused on specific editing intentions.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudnano-banana-2:officialInput 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.
Gemini 3.1 Flash Image is the model identity; the aspect ratio, resolution, quantity, and task controls above are options for this platform's invocation entry point.
You can describe goals around an existing image, such as removing reflections, replacing objects, or adjusting background tones, without having to describe the entire image again. Prompts should specify both the area to modify and the subjects, composition, and lighting to preserve, making edits more directed and suitable for product scene changes and asset updates.
Generating and modifying in-image text is a specific strength of Nano Banana 2, making it suitable for posters, menus, and interface drafts containing titles, labels, and short copy. You can include the exact copy, text placement, and layout hierarchy in the prompt, so visuals and information are designed together; final delivery should still be proofread character by character.
Use reference images to provide cues for people, products, or styles, then use text to describe the role and relationship of each asset, making it suitable for creating scene variations within the same theme. The focus is on continuity in subject recognizability and visual style, rather than generating unrelated images; key details should still be checked image by image.
Provide a product image and describe the desired desktop, indoor environment, lighting, and whitespace placement to generate scene assets for product displays or ad testing. Continue editing around the background, atmosphere, or color palette to explore different visual directions; product logos, packaging text, and appearance should be key acceptance criteria.
Provide the event theme, exact title, date, and canvas requirements to create candidate poster images with text; existing designs can also be used as references to modify the copy or seasonal atmosphere. Landscape format is suitable for display compositions, while portrait format is suitable for mobile content. The final deliverable is an image, not a design file with directly separable layers.
Provide a subject reference image and unified style instructions to generate candidate visuals for different content themes. You can set the number of generations in a single request and use asynchronous tasks and callbacks to connect with asset management workflows. When selecting, compare subject consistency, layout space, and copy accuracy, then retain versions suitable for publishing.
The base nano-banana corresponds to Gemini 2.5 Flash Image, while Nano Banana 2 corresponds to Gemini 3.1 Flash Image. If your task focuses on images with text, combining reference images, and ongoing revisions, the second generation is worth trying first. When choosing, compare results using the same product image and copy; do not interpret generational changes as meaning every image will necessarily be better.
Nano Banana 2 is suitable for repeatedly iterating generation and editing within the same workflow; nano-banana-pro is the high-quality-oriented option in the series, corresponding to Gemini 3 Pro Image. For flagship visuals, you can test both in parallel and evaluate them based on text, materials, and composition. :official is a public invocation ID variant and does not represent another independent native model.
Choose generate for text-based creation; choose edit to modify existing assets, provide reference images using image_urls, and separately describe what to preserve and what to change.
Specify model=nano-banana-2:official, action, and prompt for /nano-banana/images; start with aspect_ratio=1:1, resolution=2K, and count=1, setting the canvas format and resolution separately.
Retrieve images from data[].image_url; for asynchronous requests, query or receive callbacks using task_id. When continuing edits, pass in the selected image again and narrow the scope of changes for each round.
Create a landscape menu hero visual using the two desserts in the reference image. Write “Today's Desserts” on the left, label the two products with the given names, preserve their real-life colors, and keep the layout clear.
Check whether the products are confused, whether the names match, and whether the text is accurate; high resolution does not replace layout proofreading.
It is a public invocation ID variant of Nano Banana 2, corresponding to the Gemini 3.1 Flash Image model. Use the full ID when calling it; do not interpret :official as a new model generation, or infer additional creative capabilities or a fixed generation speed from it.
Send a POST request to /nano-banana/images, select action=edit, provide the image through image_urls, and describe the desired changes in prompt. It is recommended to also specify that the product shape and branding should be preserved and that only the background or lighting should be adjusted, so you can verify whether the result meets the requirements.
You can choose a resolution of 1K, 2K, or 4K, with seven aspect ratios including square, landscape, and portrait. First select aspect_ratio based on the publishing placement, then select resolution; these options control the output orientation, while the actual image dimensions are determined by the returned file.
You can try generating posters with Chinese titles and short copy, but you should still proofread every character, number, and punctuation mark. Provide complete and accurate copy in the prompt, and specify its position and hierarchy; when there is a lot of text or strict brand typography is required, it is recommended to treat final typesetting as a separate step.
A single request can use count to set one to four images, and you can also combine async and callback_url to organize tasks. The response includes task_id, trace_id, and image_url in the image results, making it easy to associate them with business records; after receiving them, proceed with downloading, filtering, and quality acceptance.
Choose a development approach or application suited to this model and follow its setup guide.
Python SDK Integration Guide
SDK installation, authentication and examplesTypeScript SDK Integration Guide
SDK installation, authentication and examples
