Edit

Image description prebuilt model (preview)

  • Article

[This topic is prerelease documentation and is subject to change.]

Important

  • This is a preview feature.
  • Preview features aren’t meant for production use and may have restricted functionality. These features are available before an official release so that customers can get early access and provide feedback.

The image description prebuilt model analyzes an image and generates a user-friendly description based on its analysis. It's useful in a variety of scenarios:

  • Auto-captioning: Use machine learning to generate descriptive text for images.

  • Security and surveillance: Monitor or identify threats easily with descriptions of presence, people, or activity in a video frame.

  • Search engine optimization: Describe the images in a web page to make the page more discoverable in searches.

  • Chatbots: Describe an image and reason for it, which gives the illusion of a multi-modal capability.

  • Retail environment: Describe a product to help with inventory management and restocking.

Use in Power Apps

Use the image description prebuilt model in Power Apps (preview).

Use in Power Automate

Use the image description prebuilt model in Power Automate (preview).

Licensing and role requirements

This feature is in preview. Using it doesn't deduct any AI Builder credits.

You need to have the Basic User role to use the image description prebuilt model in your apps and flows.

Supported files

  • Document formats accepted: .JPG, .JPEG, .PNG, and .BMP
  • Maximum document size: 4 MB
  • The dimensions of the image must be greater than 50 x 50 pixels

Model output

If a valid image is detected, the model tries to locate and extract the following properties:

  • Description: A description of the image in English (en-us)
  • Tags: A list of tags in the image in English (en-us)
  • Confidence score: A percentage that represents the estimated accuracy of the description