> ## Documentation Index
> Fetch the complete documentation index at: https://doc.starrise.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# GPT Image 2

> OpenAI GPT Image 2 image generation model. Next-generation autoregressive + reasoning hybrid architecture, supporting flexible resolution up to 4K with near-perfect text rendering.

GPT Image 2 is OpenAI's next-generation image generation model, available through the Starrise AI API. Built on an autoregressive + reasoning hybrid architecture, it supports native 2K resolution, near-perfect text rendering (\~99% character-level accuracy across 12+ languages), and excellent multi-object scene composition.

## Key Capabilities

* **Text-to-image** — Generate images from natural language descriptions
* **Image editing** — Edit existing images using text prompts via `/v1/images/edits`
* **Flexible resolution** — Supports arbitrary custom dimensions up to 4K (3840px), sides must be multiples of 16
* **Precise text rendering** — \~99% character-level accuracy across 12+ languages
* **Multi-object composition** — Complex scenes without occlusion or misalignment
* **Multi-style** — Photorealistic, illustration, anime, vector, 3D, data visualization

> For both `/v1/images/generations` and `/v1/images/edits`, **`n` supports 1–10, default 1**. **Requires selecting the Direct group in the console.**

## Output Specs

| Property                     | Values                                                                     |
| ---------------------------- | -------------------------------------------------------------------------- |
| Size                         | Flexible resolution (e.g., 1024x1024, 2048x2048, 3840x2160)                |
| Size constraints             | Sides: multiples of 16; aspect ratio ≤ 3:1; total pixels 655,360–8,294,400 |
| Quality                      | low, medium, high                                                          |
| Output format                | png, jpeg                                                                  |
| Input format (edit endpoint) | png, jpeg                                                                  |

## Quick Example

<CodeGroup>
  ```bash cURL theme={null}
  curl https://ai.alad.com/v1/images/generations \
    -H "Authorization: Bearer YOUR_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "gpt-image-2",
      "prompt": "A futuristic city skyline at sunset with flying cars in the sky",
      "size": "2048x2048",
      "quality": "high"
    }'
  ```

  ```python Python theme={null}
  from openai import OpenAI

  client = OpenAI(
      api_key="YOUR_API_KEY",
      base_url="https://ai.alad.com/v1"
  )

  response = client.images.generate(
      model="gpt-image-2",
      prompt="A futuristic city skyline at sunset with flying cars in the sky",
      size="2048x2048",
      quality="high"
  )

  print(response.data[0].b64_json[:100])
  ```
</CodeGroup>

## Parameters

| Parameter            | Type    | Required | Description                                                                                                                                                |
| -------------------- | ------- | -------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `model`              | string  | Yes      | Must be `gpt-image-2`                                                                                                                                      |
| `prompt`             | string  | Yes      | Image description text                                                                                                                                     |
| `n`                  | integer | No       | Number of images to generate (1–10), default `1`. **Requires selecting the Direct group in the console.**                                                  |
| `size`               | string  | No       | `{width}x{height}` format; sides must be multiples of 16; aspect ratio ≤ 3:1; total pixels 655,360–8,294,400; max single side 3,840px. Default `1024x1024` |
| `quality`            | string  | No       | `low`, `medium`, `high`, default `medium`                                                                                                                  |
| `output_format`      | string  | No       | `png`, `jpeg`, default `png`                                                                                                                               |
| `moderation`         | string  | No       | `auto` or `low`, default `auto`                                                                                                                            |
| `output_compression` | integer | No       | Compression level for jpeg format (0–100)                                                                                                                  |

## Image Editing

Edit existing images via `POST /v1/images/edits`. Request body uses multipart/form-data; images are uploaded as files.

**Up to 16 input images per request** (`image[]`).

### Input Image Formats

Supported formats: **PNG, JPEG**. Images must be submitted via **multipart/form-data file upload**.

Two available model IDs:

* `gpt-image-2` — Official model
* `gpt-image-2-c` — Cost-effective version (supports `response_format`)

> **Note:** When using `gpt-image-2-c`, selecting higher quality may affect stable output for `n` images. For stable multi-image generation, lower quality is recommended.

<CodeGroup>
  ```bash Single image theme={null}
  curl https://ai.alad.com/v1/images/edits \
    -H "Authorization: Bearer YOUR_API_KEY" \
    -F "image=@/path/to/source.jpeg" \
    -F model="gpt-image-2" \
    -F prompt="Replace the background with an ocean sunset" \
    -F n=1 \
    -F size="1024x1024"
  ```

  ```bash Multiple images (up to 16) theme={null}
  curl https://ai.alad.com/v1/images/edits \
    -H "Authorization: Bearer YOUR_API_KEY" \
    -F "image[]=@/path/to/image1.png" \
    -F "image[]=@/path/to/image2.png" \
    -F "image[]=@/path/to/image3.png" \
    -F model="gpt-image-2" \
    -F prompt="Combine the items from these reference images into a gift basket" \
    -F size="1024x1024"
  ```
</CodeGroup>

### Edit Endpoint Parameters

| Parameter            | Type       | Required | Description                                                                                                                                                                                             |
| -------------------- | ---------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `image`              | file       | Yes      | Single source image (PNG or JPEG, multipart upload)                                                                                                                                                     |
| `image[]`            | file array | Yes\*    | Multiple source images, up to **16** (PNG or JPEG). Use instead of `image` for multiple images                                                                                                          |
| `mask`               | file       | No       | Mask image with Alpha channel, same format and dimensions as source (\< 50MB). When using multiple images, the mask applies to the first image. **Requires selecting the Direct group in the console.** |
| `model`              | string     | Yes      | `gpt-image-2` or `gpt-image-2-c`                                                                                                                                                                        |
| `prompt`             | string     | Yes      | Edit instruction text                                                                                                                                                                                   |
| `n`                  | integer    | No       | Number of images to return (1–10), default `1`. **Requires selecting the Direct group in the console.**                                                                                                 |
| `size`               | string     | No       | `{width}x{height}`, default `1024x1024`                                                                                                                                                                 |
| `output_format`      | string     | No       | `png`, `jpeg`, default `png`                                                                                                                                                                            |
| `output_compression` | integer    | No       | Compression level for jpeg format (0–100)                                                                                                                                                               |
| `response_format`    | string     | No       | `url`. **Only supported by `gpt-image-2-c`.** Default `url`                                                                                                                                             |

> `image` and `image[]` are mutually exclusive and cannot be used together.

### Mask Editing

Provide a mask image with an Alpha channel to control which regions of the source image are edited. Pixels where Alpha is white (opaque) are preserved; pixels where Alpha is black (transparent) will be edited.

Requirements:

* Mask must contain an **Alpha channel**
* Must match the source image in **format and dimensions exactly**
* File size limit: \< 50MB

> Mask editing and the `n` parameter **both require selecting the Direct group in the console.**

```bash cURL theme={null}
curl https://ai.alad.com/v1/images/edits \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F image=@/path/to/source.jpeg \
  -F mask=@/path/to/mask.png \
  -F model="gpt-image-2" \
  -F prompt="Replace the background with a futuristic cityscape" \
  -F n=1 \
  -F size="2048x2048"
```

If the mask is a black-and-white image without an Alpha channel, convert it first:

```python Python theme={null}
from PIL import Image
import numpy as np

img = Image.open("bw_mask.png").convert("L")
alpha = np.array(img)
rgba = np.zeros((*alpha.shape, 4), dtype=np.uint8)
rgba[alpha > 0] = [255, 255, 255, 255]
Image.fromarray(rgba).save("mask.png")
```

<CardGroup cols={2}>
  <Card title="Generation API Reference" icon="code" href="/en/api-reference/model-api/openai/gpt-image-2">
    Interactive Playground for `POST /v1/images/generations`.
  </Card>

  <Card title="Edit API Reference" icon="pen-to-square" href="/en/api-reference/model-api/openai/gpt-image-2-edit">
    Interactive Playground for `POST /v1/images/edits`.
  </Card>
</CardGroup>