Text-in-Image Search

You can enter keywords in the search bar, and our text-in-image search using OCR technology will return images of that text. This helps you quickly find images based on the text within them regardless of the applied metadata, streamlining your upload and enrichment processes and increasing the searchability of your assets. When you upload an image to your DAM, the technology scans and extracts any text, creating fully indexed text-in-image tags. The technology can detect up to 100 words per image.

How to Enable AI Search Experience 

Reach out to your Customer Success Contact to learn how to enable the AI Search Experience in your portal and any associated costs.

Note

Any permission restrictions set on an image asset will be honored and will affect if an image asset is retrievable using this feature.

Confidence Score

A confidence level is provided for each tag, offering insights into the accuracy and reliability of the extracted text.

  1. Hover over a tag to view its confidence score.

Single-character words and special characters (e.g., © or &) are not displayed as tags.

  • Tags with high confidence scores (90% and above) are not highlighted in any color.
  • Tags with medium confidence scores (60% - 90% range) appear in the tooltip in orange.
  • Tags with low confidence scores (below 60%) appear in the tooltip in red.

output.auto.jpgScreenshot 2023-11-15 at 12.48.24 PM.png

Factors Impacting Confidence Score

  • Image Resolution: Higher image resolutions facilitate text recognition and lead to higher confidence scores.
  • Image Composition: The complexity of the background of the text can impact text recognition (i.e., neon signs, cut-off text, text on images featuring merchandise with wrinkles) and can lead to lower confidence scores.
  • Font and Text Style: Uncommon or stylized fonts may challenge text recognition. Factors like handwriting, cursive writing, or overlapping characters can lower confidence scores.
  • Language: Characters/alphabets the AI model is trained on will significantly impact confidence scores.

Supported Languages

All Bynder-supported languages that use Latin and Cyrillic alphabets, as well as Arabic, are supported. Syllabary and logographic scripts are not currently supported, and Chinese is not currently supported. 

Supported File Types

The feature works with original files in JPG or PNG format that are smaller than 15MB in size. For other formats or files exceeding 15MB, at least one derivative less than 15MB must be available on the asset. A list of supported image types can be found here. 

Not Supported

This feature will not locate text within images embedded in Word, PDF, Keynote, or PowerPoint documents.

Updated

Was this article helpful?

0 out of 0 found this helpful

We're sorry to hear that!

Find out more in our community

Need more help? Post your question in our community