How to Use Image Generation Models from OpenAI

The DALL-E image generation model dall-e-3 was retired on March 4, 2026, and is no longer available for new deployments. Existing deployments are non-functional. Use a gpt-image- series model for image generation instead. See the image generation how-to guide for updated instructions.

OpenAI’s image generation models create images from user-provided text prompts and optional images. This article explains how to use these models, configure options, and benefit from advanced image generation capabilities in Azure. You can do image generation via image generation API or responses API. Or you can experiment with image generation in the Foundry portal To select your preferred API approach and model, use the tabs at the start of this page.

Models and capabilities

Use this table to learn the differences between the different image generation models, and to help you choose the best model for your image generation needs.

Aspect	GPT-Image-2	GPT-Image-1.5	GPT-Image-1	GPT-Image-1-Mini
Availability	Public preview	Limited access preview (Apply for GPT-image-1.5 access)	Limited access preview (Apply for GPT-image-1 access)	Limited access preview (Apply for GPT-image-1 access)
Strengths	Best for high-resolution and 4K generation, improved image editing, and broad aspect-ratio support	Best for realism, instruction-following, multimodal context, and improved speed/cost	Best for realism, instruction-following, and multimodal context	Best for fast prototyping, bulk generation, or cost-sensitive use cases
Input / Output Modalities & Format	Accepts text + image inputs; outputs images only in base64 (no URL option).	Accepts text + image inputs; outputs images only in base64 (no URL option).	Accepts text + image inputs; outputs images only in base64 (no URL option).	Accepts text + image inputs; outputs images only in base64 (no URL option).
Image Sizes / Resolutions	Arbitrary resolutions: both edges must be multiples of 16 px; long edge up to 3,840 px (4K); aspect ratio up to 3:1; pixel count 655,360–8,294,400	1024×1024, 1024×1536, 1536×1024	1024×1024, 1024×1536, 1536×1024	1024×1024, 1024×1536, 1536×1024
Quality Options	Reworked quality controls: `low`, `medium`, `high`; `low` is optimized for latency-sensitive use cases	`low`, `medium`, `high` (default = high)	`low`, `medium`, `high` (default = high)	`low`, `medium`, `high` (default = medium)
Number of Images per Request	1–10 images per request (`n` parameter)	1–10 images per request (`n` parameter)	1–10 images per request (`n` parameter)	1–10 images per request (`n` parameter)
Editing (inpainting / variations)	✅ Improved editing performance with inpainting and variations	✅ Supports inpainting and variations with mask + prompt	✅ Supports inpainting and variations with mask + prompt	✅ Supports inpainting and variations with mask + prompt
Face Preservation	✅ Advanced face preservation for realistic, consistent results	✅ Advanced face preservation for realistic, consistent results	✅ Advanced face preservation for realistic, consistent results	❌ No dedicated face preservation; better for non-portrait/general creative imagery
Performance & Cost	High-fidelity, realism-optimized model; higher latency and cost	High-fidelity, realism-optimized model; improved efficiency and latency over GPT-Image-1	High-fidelity, realism-optimized model; higher latency and cost	Cost-efficient and faster for large-scale or iterative generation

Quickstart

Quotas and limits

Image generation has default rate limits per deployment:

Model	Default quota (images/min)
GPT-image-1 series	5
GPT-image-2	5

To view your current quota or request an increase, see Manage Azure OpenAI quotas.

Call the image generation API

The following command shows the most basic way to use an image model with code. If this is your first time using these models programmatically, start with the quickstart.

Image generation typically takes 10-30 seconds depending on the model, size, and quality settings.

Prerequisites

An Azure subscription. You can create one for free.
An Azure OpenAI resource created in a supported region. See Region availability.
Deploy a gpt-image-1-series or gpt-image-2 model with your Azure OpenAI resource. For more information on deployments, see Create a resource and deploy a model with Azure OpenAI.
- GPT-image-1 series models are available in limited access: Apply for GPT-image-1 access; Apply for GPT-image-1.5 access.
- GPT-image-2 is generally available (GA).
Python 3.8 or later.
- Install the required packages: pip install openai azure-identity

Send a POST request to:

https://<your_resource_name>.openai.azure.com/openai/v1/images/generations?api-version=preview

URL: Replace <your_resource_name> with the name of your Azure OpenAI resource. Required headers:

Content-Type: application/json
api-key: <your_API_key>

Body: The following is a sample request body. You specify a number of options, defined in later sections.

Set the model parameter to the name of your model deployment (for example, gpt-image-1.5).

{
    "prompt": "A multi-colored umbrella on the beach, disposable camera",
    "model": "gpt-image-1.5",
    "size": "1024x1024", 
    "n": 1,
    "quality": "high"
}

For image generation token costs, see Image tokens.

Output

The response from a successful image generation API call looks like the following example. The b64_json field contains the output image data.

{ 
    "created": 1698116662, 
    "data": [ 
        { 
            "b64_json": "<base64 image data>"
        }
    ]
} 

The response_format parameter isn’t supported for GPT-image-1 series models, which always return base64-encoded images.

Streaming

Streaming lets you receive partial images as they’re generated, providing faster visual feedback for your users. This is useful for applications where you want to show generation progress. The partial_images parameter (1-3) controls how many intermediate images are returned before the final result. You can stream image generation requests to gpt-image-1-series and gpt-image-2 models by setting the stream parameter to true, and setting the partial_images parameter to a value between 0 and 3.

import base64
from openai import OpenAI
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

token_provider = get_bearer_token_provider(
    DefaultAzureCredential(), "https://ai.azure.com/.default"
)

client = OpenAI(  
  base_url = "https://RESOURCE-NAME-HERE/openai/v1/",  
  api_key=token_provider,
)

stream = client.images.generate(
    model="gpt-image-1.5",
    prompt="A cute baby sea otter",
    n=1,
    size="1024x1024",
    stream=True,
    partial_images = 2
)

for event in stream:
    if event.type == "image_generation.partial_image":
        idx = event.partial_image_index
        image_base64 = event.b64_json
        image_bytes = base64.b64decode(image_base64)
        with open(f"river{idx}.png", "wb") as f:
            f.write(image_bytes)
 

Specify API options

The following API body parameters are available for image generation models.

Size

For GPT-image-1 series models, specify the size of the generated images as one of 1024x1024, 1024x1536, or 1536x1024. Square images are faster to generate. For gpt-image-2, arbitrary resolutions are supported with the following constraints:

Both edges must be a multiple of 16 pixels.
Long edge up to 3840 px (4K support).
Aspect ratio up to 3:1.
Total pixel count between 655,360 and 8,294,400.

Quality

There are three options for image quality: low, medium, and high. Lower quality images can be generated faster. The default value is high.

Number

You can generate between one and 10 images in a single API call. The default value is 1.

User ID

Use the user parameter to specify a unique identifier for the user making the request. This identifier is useful for tracking and monitoring usage patterns. The value can be any string, such as a user ID or email address.

Output format

Use the output_format parameter to specify the format of the generated image. Supported formats are PNG and JPEG. The default is PNG.

WEBP images aren’t supported in the Azure OpenAI in Microsoft Foundry Models.

Compression

Use the output_compression parameter to specify the compression level for the generated image. Input an integer between 0 and 100, where 0 is no compression and 100 is maximum compression. The default is 100.

Streaming

Use the stream parameter to enable streaming responses. When set to true, the API returns partial images as they’re generated. This feature provides faster visual feedback for users and improves perceived latency. Set the partial_images parameter to control how many partial images are generated (1-3).

Transparency

Set the background parameter to transparent and output_format to PNG on an image generate request to get an image with a transparent background.

Call the image edit API

The Image Edit API enables you to modify existing images based on text prompts you provide. The image editing endpoint is generally available and supported for production use. The API call is similar to the image generation API call, but you also need to provide an input image.

The input image must be less than 50 MB in size and must be a PNG or JPG file.

Send a POST request to:

https://<your_resource_name>.openai.azure.com/openai/deployments/<your_deployment_name>/images/edits?api-version=<api_version>

URL: Replace the following values:

<your_resource_name> is the name of your Azure OpenAI resource.
<your_deployment_name> is the name of your GPT-image series model deployment.
<api_version> is the version of the API you want to use. For example, 2025-04-01.

Required headers:

Content-Type: multipart/form-data
api-key: <your_API_key>

Body: The following is a sample request body. You specify a number of options, defined in later sections.

The Image Edit API takes multipart/form data, not JSON data. The example below shows sample form data that would be attached to a cURL request.

-F "image[]=@beach.png" \
-F 'prompt=Add a beach ball in the center' \
-F "model=gpt-image-1" \
-F "size=1024x1024" \
-F "n=1" \
-F "quality=high"

API response output

The response from a successful image editing API call looks like the following example. The b64_json field contains the output image data.

{ 
    "created": 1698116662, 
    "data": [ 
        { 
            "b64_json": "<base64 image data>"
        }
    ]
} 

Specify image edit API options

The following API body parameters are available for image editing models, in addition to the ones available for image generation models.

Image

The image value indicates the image file you want to edit.

Input fidelity

The input_fidelity parameter controls how much effort the model puts into matching the style and features, especially facial features, of input images. This parameter lets you make subtle edits to an image without changing unrelated areas. When you use high input fidelity, faces are preserved more accurately than in standard mode.

Input fidelity is not supported by the gpt-image-1-mini model.

Mask

The mask parameter uses the same type as the main image input parameter. It defines the area of the image that you want the model to edit, using fully transparent pixels (alpha of zero) in those areas. The mask must be a PNG file and have the same dimensions as the input image.

Streaming

Transparency

GPT-image-1 only: set the background parameter to transparent and output_format to PNG on an image generate request to get an image with a transparent background.

Write effective text-to-image prompts

Your prompts should describe the content you want to see in the image and the visual style of the image. When you write prompts, consider that the Image APIs come with a content moderation filter. If the service recognizes your prompt as harmful content, it doesn’t generate an image. For more information, see Content filtering.

For a thorough look at how you can tweak your text prompts to generate different kinds of images, see the Image prompt engineering guide.

Responsible AI and Image Generation

Azure OpenAI’s image generation models include built-in Responsible AI (RAI) protections to help ensure safe and compliant use. In addition, Azure provides input and output moderation across all image generation models, along with Azure-specific safeguards such as content filtering and abuse monitoring. These systems help detect and prevent the generation or misuse of harmful, unsafe, or policy-violating content. Customers can learn more about these safeguards and how to customize them here:

Learn more: Explore content filtering
Request customization: Apply to opt out of content filtering

Special considerations for generating images of minors

Photorealistic images of minors are blocked by default. Customers can request access to this model capability. Enterprise-tier customers are automatically approved.

Troubleshooting

API call rejection

Prompts and images are filtered based on our content policy. The API returns an error when a prompt or image is flagged. If your prompt is flagged, the error.code value in the message is set to contentFilter. Here’s an example:

{
    "created": 1698435368,
    "error":
    {
        "code": "contentFilter",
        "message": "Your task failed as a result of our safety system."
    }
}

It’s also possible that the generated image itself is filtered. In this case, the error message is set to Generated image was filtered as a result of our safety system. Here’s an example:

{
    "created": 1698435368,
    "error":
    {
        "code": "contentFilter",
        "message": "Generated image was filtered as a result of our safety system."
    }
}

Rate limit errors

If you receive a 429 error, you’ve exceeded your rate limit. Wait before retrying or request a quota increase in the Azure portal.

Authentication errors

If you receive a 401 error:

API key auth: Verify your API key is correct and not expired.
Managed identity: Ensure your identity has the Cognitive Services OpenAI User role on the resource.

Timeout errors

Image generation can take up to 60 seconds for complex prompts. If you experience timeouts:

Use streaming to get partial results sooner.
Simplify your prompt.
Try a smaller image size.

Learn about image generation tokens

How to generate embeddings with Microsoft Foundry Models service (classic)

Image prompt engineering techniques

​Models and capabilities

​Quickstart

​Quotas and limits

​Call the image generation API

​Prerequisites

​Output

​Streaming

​Specify API options

​Size

​Quality

​Number

​User ID

​Output format

​Compression

​Streaming

​Transparency

​Call the image edit API

​API response output

​Specify image edit API options

​Image

​Input fidelity

​Mask

​Streaming

​Transparency

​Write effective text-to-image prompts

​Responsible AI and Image Generation

​Special considerations for generating images of minors

​Troubleshooting

​API call rejection

​Rate limit errors

​Authentication errors

​Timeout errors

​Related content

Models and capabilities

Quickstart

Quotas and limits

Call the image generation API

Prerequisites

Output

Streaming

Specify API options

Size

Quality

Number

User ID

Output format

Compression

Streaming

Transparency

Call the image edit API

API response output

Specify image edit API options

Image

Input fidelity

Mask

Streaming

Transparency

Write effective text-to-image prompts

Responsible AI and Image Generation

Special considerations for generating images of minors

Troubleshooting

API call rejection

Rate limit errors

Authentication errors

Timeout errors

Related content