Nano Banana 2.1

Nano Banana 2.1 (Gemini Nano Banana 2.1) is a high-fidelity multimodal AI image generation and editing model developed by Google. Featuring robust subject and character consistency preservation, fine-grained instruction following, natural language understanding, and multi-reference image guidance, it enables high-quality visual content creation and local/global editing while maintaining ultra-fast generation speeds.

Base URL

https://api.icreat.ai

Authentication

All API requests require authentication using an API Key. You can obtain your API Key from the console.

export ICREAT_API_KEY="your-api-key-here"

HTTP Request Headers

import os

API_KEY = os.environ.get("ICREAT_API_KEY")
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer " + API_KEY,
}

Keep Your API Key Secure Never expose your API Key in client-side code or public code repositories. Store and use your API Key via environment variables or a backend proxy.

Code Examples

Image generation and editing follow a two-step asynchronous workflow: first submit a task to obtain a task_id, then poll the Task Result Query endpoint to check the task status. The response includes status and result. While the task is processing, result remains []; when the task status becomes SUCCEEDED, result contains the generated assets, and the response includes costUSD.

The examples below use the same task_id across the submission and polling steps.

1. Submit Task

Send an image generation or editing request to the task submission endpoint.

2. Query Task Result (Polling)

Query the task status using the task_id returned by the submission endpoint. Repeat query requests until the task reaches a terminal state.

The response includes status and result. While the task is processing, result remains []; when the status becomes SUCCEEDED, result contains the generated assets, and costUSD is included. A status of FAILED indicates that execution failed.

Input Parameters

Submit Task — Input

Total Parameters: 4, Required: 4, Optional: 0

Query Task Result — Input

Total Parameters: 1, Required: 1, Optional: 0

Output Parameters

Submit Task — Output

Total Parameters: 1

Query Task Result — Output

Total Parameters: Variable

# google/gemini-nano-banana-2-1

> Nano Banana 2.1 (Gemini Nano Banana 2.1) is a high-fidelity multimodal AI image generation and editing model developed by Google. Featuring superior subject consistency preservation, fine-grained instruction following, and multi-reference image guidance, it enables the rapid creation and editing of high-quality visual content.

## Overview

Uses the iCreat two-step asynchronous task API: first submit the image generation or editing task, then poll the task result query endpoint. Once the task succeeds, retrieve the generated image assets from `result`. Successful responses also include `costUSD`.

## API Information

- **Base URL**: `https://api.icreat.ai`
- **Submission Endpoint (POST)**: `/v1/task/submit/google/gemini-nano-banana-2-1`
- **Result Query Endpoint (POST)**: `/v1/task/result`
- **Model ID**: `google/gemini-nano-banana-2-1`
- **Authentication**: `Authorization: Bearer ${ICREAT_API_KEY}`

## Invocation Workflow

1. **Submit Task**: Send a POST request to `/v1/task/submit/google/gemini-nano-banana-2-1`. The request body includes `prompt`, `image`, `image_size`, and `aspect_ratio`. The response format is `{ "task_id": "..." }`.
2. **Query Result**: Send a POST request to `/v1/task/result` with the body `{ "task_id": "..." }`.
3. **Poll Status**: While `status` is `SUBMITTED` and `result` is `[]`, continue polling the task result.
4. **Read Result**: When `status` is `SUCCEEDED`, read `url` or `download_url` from the resource of type `Image` inside `result`.
5. **Handle Failure**: `FAILED` is a terminal state indicating task execution failure.

### Input Example

```json
{
  "prompt": "Sample multiline\ntext prompt",
  "image": [
    "[https://example.com/example.png](https://example.com/example.png)"
  ],
  "image_size": "1K",
  "aspect_ratio": "1:1"
}

Input Notes

• The request body uses top-level flat fields.
• Required fields: prompt, image, image_size, and aspect_ratio.
• prompt: Text prompt describing the image to generate or the edit operation to execute, supporting multiline text.
• image: An array of publicly accessible reference image URLs, supporting up to 4 images.
• image_size: Resolution tier, supporting 1K and 2K.
• aspect_ratio: Output image aspect ratio. Supported values include 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, and 21:9. Defaults to the aspect ratio of the first uploaded image.

Output Notes

• While the task is processing, status is typically SUBMITTED and result is [].
• Once the task succeeds, status becomes SUCCEEDED.
• The success response includes result and costUSD.
• The format of result is: [{ "type": "Image", "url": "...", "download_url": "..." }].
• FAILED is a terminal state indicating task execution failure.

Important Notes

• Each polling request must use the same task_id.
• Continue polling as long as the task is still processing and result is empty.
• Access the generated image via url, or download it using download_url.
• If the task returns FAILED, verify the prompt, image URLs, image format, aspect ratio, and other request parameters.
• Keep your API Key secure on the server side; never expose it in client-side code.