# Nano Banana 2 (nano-banana-2)

Vendor: Google
Model ID: `nano-banana-2`
Base URL: `https://api.mulerouter.ai`
Type: Inference API (async task-based)

## Description
Next-generation image generation and editing model with 4K support, 14 aspect ratios, and web search grounding

## Variant: Image to Image Edit

Endpoint: `POST /vendors/google/v1/nano-banana-2/edit`

### Input Schema
The API accepts the following input parameters:

- **`images`** (`list<string>`, _required_):
  List of reference images for editing. Supports both image URLs and base64-encoded images.  Base64 format: `data:image/png;base64,{base64_data}`  Supported formats: JPEG, JPG, PNG, BMP, WEBP Max file size per image: 20MB  Nano Banana 2 supports up to 14 reference images: - Up to 10 images of objects (high-fidelity preservation) - Up to 4 images of characters (character consistency)
  - Items: min: 1, max: 14

- **`prompt`** (`string`, _required_):
  Text prompt for image editing. Describe the desired change conversationally.  For adding/removing elements: "Add a wizard hat on the cat's head" For inpainting: "Change only the blue sofa to a brown leather chesterfield" For style transfer: "Transform this photo into Van Gogh's Starry Night style"

- **`resolution`** (`string`, _optional_):
  Output resolution preset.  **Note:** Must use uppercase K (e.g. 1K, 2K, 4K). Lowercase will be rejected.
  - Options: `"1K"`, `"2K"`, `"4K"`
  - Default: `"1K"`

- **`web_search`** (`boolean`, _optional_):
  Enable Google Search grounding for image editing. When enabled, the model can use Google Search (including Web Search and Image Search) to verify facts and generate imagery based on real-time data such as current weather, stock charts, recent events, or accurate depictions of real-world subjects.
  - Default: `false`

- **`aspect_ratio`** (`string`, _optional_):
  Aspect ratio of the edited image (width:height).
  - Options: `"1:1"`, `"1:4"`, `"1:8"`, `"2:3"`, `"3:2"`, `"3:4"`, `"4:1"`, `"4:3"`, `"4:5"`, `"5:4"`, `"8:1"`, `"9:16"`, `"16:9"`, `"21:9"`
  - Default: `"1:1"`

**Required Parameters Example**:

```json
{
  "prompt": "<Text prompt for image editing>",
  "images": []
}
```

**Full Example**:

```json
{
  "images": [],
  "prompt": "<Text prompt for image editing>",
  "resolution": "1K",
  "web_search": false,
  "aspect_ratio": "1:1"
}
```

## Variant: Text to Image Generation

Endpoint: `POST /vendors/google/v1/nano-banana-2/generation`

### Input Schema
The API accepts the following input parameters:

- **`prompt`** (`string`, _required_):
  Text prompt for image generation. For best results, describe the scene narratively rather than listing keywords.  Include details such as: - Subject (object, person, animal, scenery) - Style (photorealistic, illustration, sketch, etc.) - Lighting and mood - Camera angle and composition - Color palette  Example: "A photorealistic close-up portrait of an elderly Japanese ceramicist with deep, sun-etched wrinkles and a warm, knowing smile. Captured with an 85mm portrait lens, resulting in soft bokeh."

- **`resolution`** (`string`, _optional_):
  Output resolution preset.  **Note:** Must use uppercase K (e.g. 1K, 2K, 4K). Lowercase will be rejected.
  - Options: `"1K"`, `"2K"`, `"4K"`
  - Default: `"1K"`

- **`web_search`** (`boolean`, _optional_):
  Enable Google Search grounding for image generation. When enabled, the model can use Google Search (including Web Search and Image Search) to verify facts and generate imagery based on real-time data such as current weather, stock charts, recent events, or accurate depictions of real-world subjects.
  - Default: `false`

- **`aspect_ratio`** (`string`, _optional_):
  Aspect ratio of the generated image (width:height).
  - Options: `"1:1"`, `"1:4"`, `"1:8"`, `"2:3"`, `"3:2"`, `"3:4"`, `"4:1"`, `"4:3"`, `"4:5"`, `"5:4"`, `"8:1"`, `"9:16"`, `"16:9"`, `"21:9"`
  - Default: `"1:1"`

**Required Parameters Example**:

```json
{
  "prompt": "<Text prompt for image generation>"
}
```

**Full Example**:

```json
{
  "prompt": "<Text prompt for image generation>",
  "resolution": "1K",
  "web_search": false,
  "aspect_ratio": "1:1"
}
```

## Variant: /vendors/google/v1/nano-banana-2/edit/{task_id}

Endpoint: `POST /vendors/google/v1/nano-banana-2/edit/{task_id}`

## Variant: /vendors/google/v1/nano-banana-2/generation/{task_id}

Endpoint: `POST /vendors/google/v1/nano-banana-2/generation/{task_id}`

## Usage (Async Task API)
This model uses an async task-based workflow with two API calls:
1. **Submit a task** — `POST /v1/inference/nano-banana-2` to create a generation task
2. **Poll for result** — `GET /v1/inference/nano-banana-2/{task_id}` to check status and retrieve the result

### Step 1: Submit a Task

#### cURL
```bash
curl -X POST https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <YOUR_MULEROUTER_API_KEY>" \
  -d '{
    "prompt": "Your prompt here"
  }'
```

#### Python
```python
import requests

API_KEY = "<YOUR_MULEROUTER_API_KEY>"
ENDPOINT = "https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit"

response = requests.post(
    ENDPOINT,
    headers={
        "Content-Type": "application/json",
        "Authorization": f"Bearer {API_KEY}"
    },
    json={
        "prompt": "Your prompt here"
    }
)

result = response.json()
task_id = result["task_info"]["id"]
print(f"Task created: {task_id}")
```

#### Node.js / TypeScript
```typescript
const API_KEY = "<YOUR_MULEROUTER_API_KEY>";
const ENDPOINT = "https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit";

const response = await fetch(ENDPOINT, {
    method: "POST",
    headers: {
        "Content-Type": "application/json",
        "Authorization": `Bearer ${API_KEY}`
    },
    body: JSON.stringify({
        prompt: "Your prompt here"
    })
});

const result = await response.json();
const taskId = result.task_info.id;
console.log("Task created:", taskId);
```

#### Submit Response (202)
```json
{
  "task_info": {
    "id": "8e1e315e-b50d-4334-a231-be7d19a372f4",
    "status": "processing",
    "created_at": "2026-01-01T00:00:00.000Z"
  }
}
```

### Step 2: Poll for Result

Use the task ID from Step 1 to poll the status endpoint until the task is completed.

Endpoint: `GET /v1/inference/nano-banana-2/{task_id}`

#### cURL
```bash
curl -X GET https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit/<TASK_ID> \
  -H "Authorization: Bearer <YOUR_MULEROUTER_API_KEY>"
```

#### Python
```python
import time

status_url = f"https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit/{task_id}"

while True:
    status = requests.get(status_url, headers={
        "Authorization": f"Bearer {API_KEY}"
    }).json()
    task_status = status["task_info"]["status"]
    if task_status in ("completed", "succeeded"):
        print("Result:", status)
        break
    elif task_status == "failed":
        print("Task failed:", status)
        break
    time.sleep(5)
```

#### Node.js / TypeScript
```typescript
const statusUrl = `https://api.mulerouter.ai/vendors/google/v1/nano-banana-2/edit/${taskId}`;

while (true) {
    const statusRes = await fetch(statusUrl, {
        headers: { "Authorization": `Bearer ${API_KEY}` }
    });
    const status = await statusRes.json();
    const taskStatus = status.task_info.status;
    if (taskStatus === "completed" || taskStatus === "succeeded") {
        console.log("Result:", status);
        break;
    } else if (taskStatus === "failed") {
        console.log("Task failed:", status);
        break;
    }
    await new Promise(r => setTimeout(r, 5000));
}
```

## Additional Resources

### Documentation

- [Model Playground](https://www.mulerouter.ai/models/nano-banana-2)
- [API Documentation](https://api.mulerouter.ai/docs/api-reference/endpoint/google/nano-banana-2)

### MuleRouter Platform

- [Platform Documentation](https://www.mulerouter.ai/docs)
- [API Keys Management](https://www.mulerouter.ai/app/api-keys)