Visual Search

Cloudinary’s AI-powered Visual and Natural Language Search (referred to as Visual Search) lets you find images by describing them in plain language or by using another image as a reference. It understands context and visual meaning, whether from text or image input, to deliver more intuitive and accurate results than traditional keyword or tag-based search.

Overview

You can search visually in several ways—for example from an asset in your library, from a URL, from an uploaded file, or by describing the scene in text:

• Visual Search by Image: Allows you to find matches for a certain image in your product environment. For example, you can select an image of a woman at the top of a purple mountain, and images sharing similar traits will be retrieved.

• Visual Search by Image URL: Allows you to find matches for an image hosted on any web platform. For instance, you can input the URL of an image hosted on a website, like https://example.com/image.jpg, featuring a gold bracelet, and images in your product environment sharing similar traits will be retrieved.

• Visual Search by image upload: Allows you to upload a reference image from your computer or device when that image isn't already in your product environment and you don't have a public URL to paste. Cloudinary uses the uploaded file as the visual query to return similar images from your catalog (maximum upload size 40 MB).

• Visual Search by Text: Allows you to type in a word or phrase to find images that visually match the concept you described. For example, if you type the phrase cold night, images that visually represent that concept will be retrieved.

When you run a visual search, the images in the Media Library are scored based on how similar they are to the image you selected or the concept you typed. Matching images are then returned by score, from most to least similar.

Searching for images based on visual similarity rather than by metadata allows you to:

• Find images even if they aren't tagged or named descriptively.
• Find duplicates.
• Find images by visual characteristics that might not be included in a description.
• Gather images that are similar in appearance.
• Increase the discoverability of your images.

Searching by image

You can select any image in your product environment and find other images that are similar to it.

To search by matching image:

• From any page in the Media Library, right-click or click the (3-dots) options menu of an asset and select Find Similar Image.

URL or upload as your reference image

Use this flow when your reference image is not already stored in your product environment—for example, an image on the web (paste a URL) or a file on your computer or device (upload in the same Search by Image dialog).

To search by URL or upload:

1. Navigate to the Visual Search by clicking the Assets tab from the top of the Media Library, and clicking the Visual toggle.

   

2. Click the camera icon in the Visual Search textbox to open the Search by Image dialog.

   

3. In the Search by Image dialog, choose one of the following:

   • Paste a URL: Paste the image URL into the URL field and click Search.
   • Upload a file: Drag an image into the upload area, or click Upload a file and select a reference image from your computer or device (maximum 40 MB), then click Search.

Searching by text

You can type in a search text to find images that are visual similar to the concept you enter.

To search by text:

1. Navigate to the Visual Search by clicking the Assets tab from the top of the Media Library, and clicking the Visual toggle.

2. Enter your search word or phrase and see the matching results: