nano-banana-2-lite corresponds to Google's Gemini 3.1 Flash Lite Image, an image model in the Nano Banana series focused on speed and cost efficiency. It handles text-to-image generation and image editing with native 1K output, making it suitable for creative drafts, asset variations, and high-frequency visual tasks; for high-resolution final work or complex continuous editing, you can choose other models in the series.
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.
1K is this model's native resolution. Aspect ratios and return formats vary by endpoint, and general parameters do not mean every capability applies to Lite.
Lite is designed with speed and cost efficiency in mind, making it suitable for turning clear visual descriptions into images for review. Prompts can be structured around the subject, scene, lighting, composition, and style: explore different creative directions first, then select options worth developing further, rather than taking on complex final-production tasks from the outset.
In addition to generating from scratch, you can also provide an image and editing instructions to try changing the background, adjusting colors, or altering the visual style. In practice, clearly specify both what needs to change and what needs to remain, and keep each task focused on a single goal whenever possible; Lite is better suited to lightweight edits than to consistency that depends on long-term continuous editing.
The dedicated endpoint provides square, landscape, and portrait aspect ratios, making it easy to organize compositions around covers, content illustrations, and mobile display. Decide the purpose and subject placement first, then describe negative space, viewpoint, and background relationships; this is usually more appropriate than forcing a crop after generation. Different aspect ratios still use 1K output and do not automatically increase the level of detail.
Enter the event theme, colors, subject, and atmosphere to generate cover or key visual drafts for internal discussion. You can separately try photographic, illustrated, and minimalist compositions, delivering candidate images and corresponding prompts. After confirming the direction, assign tasks requiring refined text, brand guidelines, or large-format output to a more suitable model.
Provide clear product images and request desktop, indoor, or outdoor backgrounds, or explore different color schemes. The deliverables are suitable as concept previews for product visuals, making it easier to compare scenes and atmospheres. Prompts should emphasize the contours and structure to preserve, and packaging, labels, and key product details should be checked one by one before formal use.
Combine the article topic, section style, and presentation direction into prompts to generate image candidates for content editing tools. Landscape images are for article headers, portrait images are for mobile covers, and square images are for cards. The application can save image links and task identifiers so editors can select results instead of treating a single generation directly as the final publishable version.
If tasks mainly involve standalone generation, simple editing, and 1K assets, Lite better fits high-frequency creation. Nano Banana 2 corresponds to Gemini 3.1 Flash Image, targeting more general visual tasks, supporting higher resolutions, and emphasizing multi-reference image processing and consistency. When multiple assets need to be combined or continuous revisions are required, prioritize the standard version rather than comparing names alone.
Nano Banana Pro corresponds to Gemini 3 Pro Image and is positioned for complex visual tasks, brand consistency, and refined creative control. Lite can handle early-stage direction exploration, but it should not replace every refinement step. If delivery requirements include strict brand elements, complex layouts, or professional visual final artwork, choose Pro based on these requirements and reserve time for final review.
Choose generate to create from text; choose edit to modify existing assets, provide reference images with image_urls, and separately describe what to preserve and what to change.
Specify model=nano-banana-2-lite, action, and prompt for /nano-banana/images; start with aspect_ratio=1:1, resolution=1K, and count=1; Lite does not support 2K/4K.
Get the image from data[].image_url; for asynchronous requests, query with task_id or receive a callback. When continuing edits, pass in the selected image again and narrow the scope of changes in each round.
Generate a set of concise product image concept drafts with a consistent style: a green glass water cup on a creamy white tabletop, front view, soft shadows, with no text in the image.
Use 1K to validate composition and color scheme; after selecting a direction, compare the standard version or Pro if higher resolution or complex continuous editing is needed.
No. Lite corresponds to Gemini 3.1 Flash Lite Image, while the original Nano Banana corresponds to Gemini 2.5 Flash Image. When making calls, you should explicitly specify nano-banana-2-lite; do not omit the model name simply because both belong to the Nano Banana service, or assume the two versions have exactly the same capabilities.
2K and 4K cannot be regarded as Lite's native output capabilities; its only resolution tier is 1K. If a project requires higher resolution, consider Nano Banana 2 or Pro. Changing the aspect ratio and selecting a higher resolution are different operations; portrait or widescreen compositions do not mean more detail.
When using /nano-banana/images, select action=edit, provide images through image_urls, and enter editing instructions; when using /openai/images/edits, provide image and prompt, and explicitly specify the model. It is recommended to complete focused edits on a clear original image first, then assess whether to continue editing.
You can try simple edits, but Lite is not optimized for multi-turn continuous editing, so you should not expect identity and details to be strictly preserved in every round. When you need to continuously maintain a person's appearance, product structure, or relationships among multiple reference assets, the standard Nano Banana 2 is better suited to such tasks, and each round should still be reviewed.
Images from the dedicated endpoint are located in data[].image_url; the OpenAI image endpoint result structure includes URL or Base64 data and also defines a task ID return format. Applications should distinguish between image results and task identifiers and save necessary records; all generated images contain a SynthID watermark.
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
