Text is optional in a scan
DjVu can store an OCR transcription alongside its compressed image. This product reads that transcription; it does not recognize new text from pixels. An empty text sidecar can be a correct result.
Visible image and hidden text differ
The page image is the visible evidence. Existing OCR can contain errors even when the image looks correct. Verify important names and numbers against the page itself.
Font and rotation limits are reported
Supported glyphs on upright pages are placed in an invisible PDF text layer. Other extracted text remains available as UTF-8. This avoids silently inserting replacement characters as if they were accurate.
Test a phrase before relying on extraction
Choose a page containing a distinctive phrase such as a chapter title. Convert only that page, open the PDF in a viewer with search, and search for the phrase. Then open the UTF-8 text download and compare the same words with the visible scan. These checks answer different questions: whether text was copied and whether the old transcription was accurate.
The displayed searchable-zone count records zones written into the PDF, not recognized words or an accuracy score. A high count does not repair an incorrect OCR layer. A zero count can be expected for an image-only source, a page without text zones, or pages whose existing text cannot be placed with the supported font and orientation.
For example, a two-page selection may contain one upright English page and one rotated page. Both images can appear correctly in the PDF while only the first contributes a searchable layer. The second page's existing transcription stays in the text sidecar, and the result explains that limitation rather than concealing it.
Copying existing OCR is useful for reference and discovery, but names, dates and account numbers still need visual checking. If fresh recognition is required, run a separate OCR application on a copy of the completed PDF. Retain the original DjVu and the unedited sidecar so you can distinguish original text from later OCR changes.