Skip to content

🧩 AI-102 Study Card – Image Generation with Azure AI Foundry (DALL-E)

✅ 1. Problem

You need to generate original images (not from catalogs) based on natural language descriptions. Real-world examples: - Illustrations for articles or marketing campaigns - Creating logos or product graphics - Unique content for websites, e-commerce, social media

🧠 2. Solution in Azure

Use Azure AI Foundry with generative AI models like DALL-E 3 or GPT-Image 1.

✅ These models allow you to: - Generate images from text prompts (e.g., "a squirrel on a motorcycle") - Customize output in terms of resolution, style, and quality - Obtain original and synthetic images, not searched online

🔩 3. Required Components and Tools

Component Description
Azure AI Foundry Platform that hosts and orchestrates generative models
DALL-E 3 Specific model for image generation
REST API or SDK To make programmatic requests
Azure AI Foundry Playground UI to test prompts and generate images

🧱 4. Architecture/Development

  • Input: Text prompt (e.g., "A badger wearing a tuxedo")
  • API Call: POST with JSON payload
  • Output: URL of generated image in PNG format

Example JSON Request:

{
  "prompt": "A badger wearing a tuxedo",
  "n": 1,
  "size": "1024x1024",
  "quality": "hd",
  "style": "vivid"
}

Response:

{
  "data": [
    {
      "url": "<Generated image URL>",
      "revised_prompt": "<System-rewritten prompt>"
    }
  ]
}

💡 5. Best Practices / Considerations

  • DALL-E 3 only supports n = 1
  • Supported resolutions: 1024x1024, 1792x1024, 1024x1792
  • Quality: standard or hd (default: standard)
  • Style: vivid or natural (default: vivid)
  • Prompt automatically improved by the system
  • Output is synchronous and directly provides the image URL

Additional Technical Details:

API Endpoint Format:

POST https://<your-resource-name>.openai.azure.com/openai/deployments/<deployment-name>/images/generations?api-version=2024-02-15-preview

Required Headers:

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

Complete Code Example (Python):

import requests
import json
import os

# Configuration
endpoint = "https://<your-resource>.openai.azure.com"
api_key = "<your-api-key>"
deployment_name = "<your-dalle-deployment>"

# API request
url = f"{endpoint}/openai/deployments/{deployment_name}/images/generations?api-version=2024-02-15-preview"

headers = {
    "Content-Type": "application/json",
    "api-key": api_key
}

payload = {
    "prompt": "A futuristic city with flying cars at sunset",
    "n": 1,
    "size": "1024x1024",
    "quality": "hd",
    "style": "vivid"
}

response = requests.post(url, headers=headers, json=payload)
result = response.json()

# Get the generated image URL
image_url = result["data"][0]["url"]
print(f"Generated image URL: {image_url}")

Complete Code Example (C#):

using Azure;
using Azure.AI.OpenAI;

// Initialize client
var client = new AzureOpenAIClient(
    new Uri("https://<your-resource>.openai.azure.com"),
    new AzureKeyCredential("<your-api-key>")
);

// Generate image
var imageOptions = new ImageGenerationOptions()
{
    Prompt = "A futuristic city with flying cars at sunset",
    Size = ImageSize.Size1024x1024,
    Quality = ImageGenerationQuality.High,
    Style = ImageGenerationStyle.Vivid
};

Response<ImageGenerations> response = await client.GetImageGenerationsAsync(
    "<your-dalle-deployment>", 
    imageOptions
);

// Get the generated image URL
string imageUrl = response.Value.Data[0].Url.ToString();
Console.WriteLine($"Generated image URL: {imageUrl}");

🔧 Configuration Parameters

Parameter Type Options Description
prompt string Any descriptive text The description of the image to generate
n integer 1 (only supported value) Number of images to generate
size string 1024x1024, 1792x1024, 1024x1792 Resolution of the generated image
quality string standard, hd Quality level of the image
style string vivid, natural Style of the generated image

📊 Pricing and Limitations

  • DALL-E 3 pricing is based on image resolution and quality
  • Rate limits apply per deployment
  • Images are temporarily stored and URLs expire after a certain period
  • Content filtering is applied to both prompts and generated images

❓ 6. Simulated Exam Questions

  1. 🔹 You need to generate images from text. Which Azure service do you use?
    ➜ Azure AI Foundry + DALL-E 3

  2. 🔹 What parameters can you set in the payload to control image quality?
    ➜ size, quality, style, prompt, n

  3. 🔹 How do you get the result from DALL-E 3 via API?
    ➜ The URL of the generated image is returned in the JSON response.

  4. 🔹 What are the supported image resolutions for DALL-E 3?
    ➜ 1024x1024, 1792x1024, 1024x1792

  5. 🔹 Can you generate multiple images in a single API call with DALL-E 3?
    ➜ No, DALL-E 3 only supports n = 1

  6. 🔹 What happens to the original prompt when using DALL-E 3?
    ➜ The system automatically improves and rewrites the prompt for better results

🚀 Quick Start Checklist

  • [ ] Create Azure AI Foundry resource
  • [ ] Deploy DALL-E 3 model
  • [ ] Get endpoint and API key
  • [ ] Test in Azure AI Foundry Playground
  • [ ] Implement via REST API or SDK
  • [ ] Handle image URL response
  • [ ] Implement proper error handling
  • [ ] Consider content filtering policies