About
OCR (Optical Character Recognition) makes it easy to search for text within a video, like titles, slide content, labels on graphs, or other on-screen text. This helps students and employees quickly find what they’re looking for, improving productivity and saving time.
This article covers how to order OCR and view OCR requests in Kaltura’s Video Portal.
For OCR functionality to work, Reach services must be enabled on the account.
- Want OCR to be ordered automatically? Check out our article Create automated REACH rules for channels and galleries.
- To learn about OCR results in search, check out the Global Search article.
Access the captions & enrich service
- Go to the media page for the desired video and click the three dots below the player to open the menu.
- Select Reach captions & enrich from the menu.
The 'Reach captions & enrich' screen displays.
Order OCR
- Click Order (bottom right).
- In the Feature drop-down menu, select OCR.
- Click Order (bottom right).
A confirmation message displays: Your request has been received. Your video will automatically be updated upon completion.
Click Back to view your order as well as any other existing orders.
Request status types
OCR requests can have the following statuses:
- Pending - The request is being created
- Processing - Sent to the vendor and the vendor received the request
- Ready - OCR is complete
- Error - The request failed; error details are included
- Aborted - The request was canceled
Hover over the information icon (i) next to the request to display details.
Cancel a request
You can cancel pending requests by clicking the trash can icon next to the request.
A confirmation message displays. Click 'Yes' to confirm.
If you try to delete a request that's already in progress, a message displays: The request is already in progress, therefore cannot be canceled.
OCR language support
- OCR is optimized for Latin-based characters, numbers, and common symbols (e.g., @, #, $, %).
- Non-Latin scripts may be partially supported:
- Chinese, Japanese, Korean: Works best with printed text in standard fonts.
- Arabic, Hindi, and others: Accuracy may vary and is not explicitly supported.