How long does it take?

Roughly 30 seconds per page on the web app, faster via the API batch endpoint. A 1,000-page archive = ~8 hours via the API, plenty parallel-able.

Will the file size grow?

Yes — by the size of the text layer. A 10 MB scan becomes ~10.5 MB after OCR. Negligible.

30+ scripts: Latin, Cyrillic, Greek, CJK (Chinese/Japanese/Korean), Arabic, Hebrew, Devanagari, Bengali, Tamil, Thai, more. Setting the language explicitly improves accuracy.

Use case · Depends on volume — 30 seconds per page, automatable

How to OCR a folder of scanned PDFs so they become searchable

Make decades of scanned archives findable via Cmd+F, Spotlight, and Google Drive search.

For: Archivists, records managers, lawyers digitizing old case files, anyone with a scanner-in-a-box

Start with step 1 Browse all tools

STEP BY STEP

The workflow

Pre-process: deskew and clean up scans

Tilted scans OCR poorly. Run our auto-deskew first, especially on flatbed scans where pages drift. The OCR step itself includes a basic deskew, but pre-processing improves accuracy on borderline scans.

Open OCR & Smart Extract

Run AI OCR on every page

Upload one file at a time, or use the batch endpoint via the API. Each page becomes searchable. Multilingual? Set the language explicitly for cleaner results on mixed-language documents.

Open OCR & Smart Extract

Save as searchable PDF (looks identical)

The output is the same scan visually, but with a hidden text layer. Spotlight, Windows Search, Google Drive, and SharePoint all index it. Cmd+F works in any reader.

WHY IT MATTERS

The case for doing this

An archive you can't search is a haystack with no needle. Most legacy archives — medical records, old contracts, court files, family genealogy — sit as image-only PDFs because that's what scanners produced. OCR is the bridge between 'we have it somewhere' and 'we can find it in 10 seconds.' For lawyers, this can be the difference between finding a smoking-gun email in 30 minutes and 30 hours.

COMMON MISTAKES

What to avoid

OCR'ing low-resolution scans

Below 200 DPI, accuracy drops fast. If you can rescan at 300 DPI, do — the OCR savings outweigh the rescan cost.

Skipping language specification

Auto-detect works most of the time, but for mixed-language archives (Spanish/English law firm, French/Dutch corporate), explicitly setting both languages improves accuracy by several percent.

Trusting OCR'd numbers without verification

0/O, 1/l, 5/S confusions are real. For dollar amounts, dates, account numbers in legal contexts, manually verify samples.

TIPS FOR THE BEST RESULTS

What works

Run on copies, not originals

Always preserve the unsearchable original alongside the searchable output. Re-OCR is cheap; re-scan is expensive.

Output as PDF/A-2u for archival

PDF/A-2u is the searchable archival ISO standard — required for many regulatory archives. Toggle it on for compliance use cases.

Save as a Macro for repeat workflows

If you OCR a folder every Monday, save the deskew + OCR + searchable-PDF chain as a Macro. Automation pays back quickly.

RELATED WORKFLOWS

Other jobs people do with pdfcraft ai

How to extract every table from a financial PDF into one spreadsheet

10-K, 10-Q, annual report → clean Excel with one sheet per table, ready to chart.

How to combine bank statements into one PDF for your accountant

Stitch 12 monthly statements into one searchable, OCR'd PDF in under 5 minutes.

How to redact a PDF properly before sending it externally

Permanently remove names, salaries, account numbers, and PII — not 'cover with a black box'.

FAQ

Questions about this workflow

On 300 DPI grayscale scans of typed documents: 98%+. On 200 DPI scans: 95%+. On photo-quality scans of typed text: 96%+. On handwriting: low — use the AI handwriting model for those, accuracy depends heavily on penmanship.

Try the workflow on a real PDF

Free tools run in your browser. No signup, no watermarks.

Start with step 1