CertPrep Coach

Vision and image generation

Use multimodal models to interpret images, generate new images from prompts, and build lightweight apps with vision capabilities.

Three vision paths Images can go into a multimodal model together with a text prompt to get answers about them. A text prompt can go into an image generation model to create a new image. Azure Vision returns structured analysis such as tags, objects and text. Image + question “What’s damaged in this photo?” Multimodal model reads the image and the prompt together Answer description, details, reasoning Text prompt “A product photo on a white background” Image generation model creates or edits images New image generated output Image photo or scan Azure Vision classic image analysis Structured results tags, objects, captions, printed text
Images can go into a multimodal model together with a text prompt to get answers about them. A text prompt can go into an image generation model to create a new image. Azure Vision returns structured analysis such as tags, objects and text.

What Microsoft lists for this area

  • Interpret visual input in prompts by using a deployed multimodal model
  • Create new visual outputs by using generative models
  • Build a lightweight application that includes vision capabilities

What this area tests

  • interpreting visual input in prompts with a deployed multimodal model
  • creating new visual outputs with generative models
  • building a lightweight app that includes vision capabilities

Images in prompts

A multimodal model accepts images alongside text. You send the image together with a question, such as “List any visible damage on this car,” and the model answers in text. This handles open-ended questions well: describing a scene, comparing two pictures, reading a chart, or explaining what’s wrong in a screenshot.

Tips that show up in scenarios:

  • Be specific about what you want from the image, and in what format.
  • For structured results, ask for JSON with named fields.
  • Image quality matters: blurry or cropped images give weaker answers.

Generating images

Image generation models create new images from a text description, and some can edit an existing image. Describe the subject, style, setting and anything to avoid. Generated images are subject to content filters, which connects back to responsible AI.

Azure Vision

Azure Vision in Foundry Tools offers classic image analysis: it extracts visual features such as objects, tags and auto-generated captions, and reads printed text. It returns consistent, structured results without prompts. Microsoft has announced that the Image Analysis 4.0 service will be retired in 2028 and points to alternatives, so for new apps, multimodal models and Content Understanding are often the better fit.

Which approach fits?

  • Open-ended questions about an image: a multimodal model.
  • Creating marketing visuals or illustrations: an image generation model.
  • Pulling specific fields from scanned forms or IDs: information extraction with Content Understanding (see that guide).

Practice Vision and image generation

19 practice questions with explanations for this area.

Practice this area Or read 12 sample questions with answers

Sources

Updated October 25, 2026. Microsoft changes products often; check the official study guide for the latest outline.