Convert Scanned PDF to Searchable PDF

How it works

OCR review checklist

Check names, dates, totals, decimals, and page headings against the scan. Confidence is a warning signal, not proof that every character or column was interpreted correctly.

The browser opens the source with pdf.js and renders each selected page at a fixed scale. Tesseract receives the rendered canvas and returns page text, word confidence, and bounding boxes. pdf-lib then opens the original document and writes each recognized word at the corresponding PDF coordinate. The text is drawn with zero opacity so the scan remains the visual source while PDF search and extraction can see the word layer.

The coordinate conversion matters: a canvas starts at the top left while PDF page coordinates start at the bottom left. Each OCR bounding box is therefore scaled back from the rendered pixels into PDF points and its bottom edge is measured upward from the page height. Unsupported glyphs may be omitted from the overlay because the local built-in font has a limited character set; the separate TXT download retains the OCR engine's text.

Prepare a scan for better OCR

When to use it

OCR works best when the page is straight, evenly lit, and large enough that small letters have several pixels of height. Crop dark borders, rotate a skewed scan, and avoid glare or a shadow crossing the text. Printed black text on a light page usually gives a more useful layer than faint carbon copies, decorative type, or a page with several columns and handwritten annotations. If a PDF contains both scanned pages and native text, process a copy and compare the searchable result with the original. Smaller batches keep memory use predictable and make it easier to review confidence and page-level errors before the document becomes part of a workflow.

Use OCR PDF for archived forms, receipts, invoices, printed notes, and scans that need basic search or copy support. Review names, numbers, dates, totals, checkboxes, and columns manually. It is not a redaction tool, a handwriting transcription service, a translation engine, or a certified accessibility remediation workflow. Preserve the source scan and treat the generated text as an editable interpretation.