Introducing AzureImageSDK — A Unified C# SDK for Azure Image Generation & Captioning, Analysis & Packed With Powerful Utilities.

[DrHazemAli]

Introduction

I'm excited to introduce AzureImageSDK — a comprehensive and modern .NET SDK for integrating Azure AI Foundry's image models, including Stable Image Ultra, Stable Image Core, Azure Vision, and moderation APIs — all in a single, extensible library.

While building solutions using Azure's AI image services, I noticed a gap: No unified, clean, and developer-friendly SDK for image generation, vision analysis, and content moderation—all in one package.

So I created AzureImageSDK — a fully open-source C# SDK that wraps Azure’s powerful image tools into a single, async-friendly interface.

The goal?
Make it simple to turn prompts into images, analyze visual content, and handle moderation, with just a few lines of code.

Features and Screenshots

🔥 Highlights

The SDK is designed for scale, its extendable architecture allows adding more models as soon as it gets released.

✅ Image AI Clients

• GPT-Image-1
• DALL-E-3
• Stable Image Ultra/Core generation
  • Stable Image Ultra
  • Stable Image Core
• Azure Vision captioning (single & dense)
• Image to Text (Image Captioning)

⚙️ Utility Modules (all async-ready)

• ImageFormatConverter – Convert between MIME types & extensions, validate formats
• ImageMetadataExtractor – Extract dimensions, file size, format, timestamps, properties
• ImageQualityAnalyzer – Analyze sharpness, brightness, contrast, overall score
• ImageSizeValidator – Validate dimensions, enforce constraints, scale to fit/fill
• AspectRatioConverter – Convert and validate aspect ratios like "16:9" to (width, height)

Technical Details

🛠️ Technical Details

🧪 Built With:

• .NET 6.0+ / .NET 8.0
• C# 10 / 11 (async/await, pattern matching, records)
• Azure AI services: OpenAI, Vision, Content Safety
• REST API integration using HttpClientFactory
• JSON serialization via System.Text.Json (and optional fallback to Newtonsoft)

---

🏗️ SDK Architecture:

• Modular Design: Each AI capability (image gen, captioning, moderation) is in a dedicated model layer
• Generic Request/Response Pipeline: Easily adaptable to any new Azure model by plugging in request/response DTOs
• Utility Toolkit: Packaged separately under AzureImage.Utilities.* namespace
• Fully Async: All operations are Task-based for non-blocking IO

---

🧩 Key Modules:

Module	Purpose
StableImageUltraModel, StableImageCoreModel	AI-powered image generation with parameter-rich configs
GPTImage1Model	AI-powered image generation with parameter-rich configs (Generating & Editing)
DALLE3Model	AI-powered image generation with parameter-rich configs
AzureVisionCaptioningModel	Captioning and dense captioning via Azure Vision API
AzureContentModerationClient	Image safety score detection
ImageFormatConverter	Convert and validate between formats and MIME types
ImageMetadataExtractor	Pull metadata from image files or streams
ImageQualityAnalyzer	Sharpness, brightness, contrast scoring
ImageSizeValidator	Validate & scale image dimensions
AspectRatioConverter	Convert and validate aspect ratio strings like "16:9"

---

🛡️ Reliability & Resilience:

• Built-in retry policy with exponential backoff using Polly
• DI-ready for ASP.NET Core and generic .NET hosts
• Configurable via appsettings.json, environment vars, or code
• Comprehensive exceptions: Rich error types with error codes, status codes, and resolution hints

---

📈 Extensibility:

• Add support for new Azure models by:
  • Creating a new model class and DTOs
  • Plugging into the generic AzureImageClient.GenerateImageAsync<TReq, TResp>()
• Drop-in support for multi-service environments via scoped dependency injection

Challenges and Solutions

💡 The SDK Solves Big Gaps in the Azure Image AI Ecosystem

🧩 Problem Area	✅ What I Built
No native .NET support for Azure image AI (Stable Image Ultra, Vision, Content Safety)	Unified, fully async C# SDK with typed models
No native or unified architecture for new Azure models	A modular SDK architecture that easily adapts to new model APIs
No utilities for validating or processing image files for AI workflows	5 robust utility modules for formats, metadata, quality, sizing, and aspect ratios

---

🔧 Other Real-World Challenges I Solved

While working hands-on with Azure image services, I hit multiple developer friction points. Here’s what AzureImageSDK addresses directly:

🛠️ Challenge	✅ AzureImageSDK Fix
Multiple APIs with inconsistent formats	Unified, modular C# clients with consistent patterns
Manual payload crafting and HTTP calls	Developer-friendly async methods like GenerateImageAsync(...)
Missing .NET support for Dense Captioning	Strongly typed wrappers for captioning, bounding boxes, and regions
Retry logic is hard to manage	Built-in exponential backoff with customizable policies
No built-in image processing or validation tools	Integrated utilities for MIME conversion, quality scoring, and size validation