Azure AI Image Analysis and Generation Resource/FAQ

Azure Vision Service
Analyzes images and video, and can also extract data from them. Azure Vision is a standalone service within the Azure portal. To use it, simply create a new Azure Vision service in the portal.

Creating New Images
Image generation is handled under Azure OpenAI. You will need to create a new Azure OpenAI service and select one of the image models available in the Azure model catalog. There are many options to choose from; at the time of this writing, I am seeing 13 image models available in the East US region when using Foundry Playground (more below).

Because not all models can generate images - only specific multimodal ones can - you will need to test different image models to find the one that best suits your needs. You can use the Microsoft Foundry Playground to test these models interactively.

Note: the image playground sometimes errors out, but often times, it still creates the deployment. Make sure you check the "Deployments" drop-down menu, after the error, to see if it is there.

It can also take a minute or two for the image to generate. I have seen it respond in a few seconds and other times when it took a minute or two.

APIs & Containers
Both services can be integrated with a custom API once established.

Azure Vision service can be containerized. So, if you don't want to send your photos out over the wire (via an API) then you can take advantage of Azure Vision service inside an Azure AI Container to handle data locally. This is also a good way to avoid latency if your files are large.

Conversely, Azure OpenAI does not support local container deployments; it must remain cloud-based due to the required computing architecture.