Image Editing with Azure OpenAI GPT-image-1

This repo demonstrates how to use Azure OpenAI GPT-image-1, a powerful image generation model in Azure AI Foundry. The provided Jupyter notebook showcases image editing capabilities, allowing you to modify existing images based on textual prompts.

This demo leverages the requests library for direct API interaction and azure-identity library for secure Azure Entra ID authentication.

---

🎧 You may listen to the audio podcast, "Image Makeovers with GPT-image-1: Behind the Code", where we explore the content of this notebook in a beginner-friendly way.

---

📑 Table of Contents:

• Part 1: Configuring Solution Environment
• Part 2: Performing Image Edits and Visualisation

---

Part 1: Configuring Solution Environment

To use the notebook, set up your Azure OpenAI environment and install Python packages.

1.1 Azure OpenAI Service Setup

Ensure you have an Azure OpenAI Service resource with the GPT-image-1 model deployed.

1.2 Authentication

This demo uses Azure Entra ID authentication via DefaultAzureCredential from azure.identity. This credential type automatically handles various authentication methods.

To enable this, ensure your environment is set up for Azure authentication, e.g., by logging in via az login (Azure CLI) or setting relevant environment variables for service principals.

1.3 Environment Variables

Set the following environment variables, pointing to your Azure OpenAI GPT-image-1 deployment:

Environment Variable	Description
AOAI_API_BASE	Your Azure OpenAI endpoint URL (e.g., https://<YOUR_AOAI_RESOURCE>.openai.azure.com).
AOAI_DEPLOYMENT_NAME	The name of your GPT-image-1 deployment (e.g., gpt-image-1).
AOAI_API_VERSION	The API version for image edits (e.g., 2025-04-01-preview).

1.4 Install Required Python packages

pip install requests Pillow ipython azure-identity

Part 2: Performing Image Edits and Visualisation

The GPT-image-1.ipynb notebook outlines the core image editing logic:

• API URL Construction: Builds the endpoint URL using environment variables.
• Input Image: Loads the image from INPUT_IMAGE_PATH (e.g., ./input_image.jpeg).
• Secure Authentication: Obtains an access token using DefaultAzureCredential.
• API Request: Sends a multipart/form-data POST request to the GPT-image-1 API, including:
  • The input image file.
  • A prompt string (e.g., "Add a giant, friendly robot behind the bird.").
  • Parameters like n (number of images, typically 1) and size (e.g., 1024x1024), along with other options.
• Response Processing: On success, it decodes the base64-encoded edited image, saves it to OUTPUT_IMAGE_PATH (e.g., ./output_image_edited.png), and displays it.
• Error Handling: Includes robust error reporting for API failures.

Example API call structure from the notebook:

# Input image
files = {
    "image": (
        os.path.basename(INPUT_IMAGE_PATH),
        image_file,
        "image/jpeg")
}

# Input prompt and parameters
data = {
    "prompt": PROMPT,
    "n": 1,
    "size": "1024x1024",
    "quality": "medium",
    "output_compression": 100,
    "output_format": "jpeg"
}

# Entra ID Auth
credential = DefaultAzureCredential()
token = credential.get_token("https://cognitiveservices.azure.com/.default")
headers = {
    "Authorization": f"Bearer {token.token}"
}

# Image API request
print("Making API call to edit the image...")
response = requests.post(
    API_URL,
    headers = headers,
    files = files,
    data = data
)

Note

Feel free to modify the PROMPT string to experiment with different image editing requests.

Example Input:

This is an example of an input image that could be used for editing in this notebook.

Example Output:

Below is an example of an edited image, based on an input image and a prompt like "Add a giant, friendly robot behind the bird."

Note

Your actual output may vary.