Tiny Online Tools logoTiny Online ToolssearchSearch tools…grid_viewAll Tools
Homechevron_rightPDF Toolschevron_rightPDF OCR CorrectorPDF OCR Corrector

PDF OCR Corrector

Fix misrecognised words in a scanned PDF's hidden OCR layer, leaving the scan pixel-identical.

upload_file

Click to browse or drag & drop files here

Select a scanned PDF with an OCR layer

Accepted: .pdf,application/pdf

Similar Tools

PDF OCR Layer Viewer

PDF OCR Layer Viewer

See the invisible OCR text of a scanned PDF drawn exactly where it sits on the page.

Scanned vs Searchable PDF Detector

Scanned vs Searchable PDF Detector

Tell real text from a bare scan and from a scan with an OCR layer, page by page.

PDF Hidden Data Inspector

PDF Hidden Data Inspector

Reveal what a PDF carries but does not show: metadata, hidden layers, attachments, JavaScript, invisible text and text under redaction boxes.

PDF Metadata Editor

PDF Metadata Editor

View and edit PDF metadata including title, author, subject, keywords, and creator.

PDF Form Data Extractor

PDF Form Data Extractor

Export filled PDF form values to JSON, CSV, FDF or XFDF — including batch mode: many forms in, one spreadsheet out.

Video Frame Extractor

Video Frame Extractor

Extract individual frames from a video as PNG images at any interval, entirely in your browser.

PDF Annotation Remover

PDF Annotation Remover

Remove or flatten comments, highlights and review markup in a PDF — filtered by type, author, date or page, with a count of exactly what will go.

apps

More Tools

Browse our full collection of free online tools.

Correcting what nobody can see

A scanned PDF that has been through OCR carries the scan as an image and the recognised words as invisible text sitting on top of it. When the recogniser gets a word wrong, the picture still looks perfect — it is only search, copy-paste and screen readers that quietly fail.

This tool edits that hidden layer. The scan is not re-encoded, re-compressed or re-drawn.

How the file is changed

The page's content stream is a list of operators: some draw the scanned image, some position and show text. The stream is lexed, and only the string operand of an invisible text operator is replaced. Every other byte's meaning is preserved — the positioning operators, the graphics state, and the operator that paints the scan all pass through untouched.

That is why the corrected page cannot shift, blur or recompress. It is not a promise; it is a consequence of not regenerating anything.

Verification you can read

After writing, the result is re-opened and three checks run on every changed page:

  1. Text — the corrected words are what a text extractor now returns.
  2. Invisibility — the layer is still drawn in render mode 3, and the count of visible glyphs is unchanged.
  3. Pixels — the corrected page is rendered and compared against the original render, pixel by pixel. The changed ratio should be zero.

All three results are shown before you download, pass or fail.

When it refuses

Some pages cannot be rewritten safely: a content stream using a filter that cannot be decoded, or a text block that mixes visible and invisible glyphs so the visible ones cannot be guaranteed intact. Those pages are refused by name and left exactly as they were. A refusal is the right answer — a silently corrupted file is not.

Characters the PDF's fonts cannot express are handled the same way: where possible a standard Helvetica is embedded to carry the replacement, and where that is not enough, the edit is skipped and named.

Privacy

Parsing, editing, saving and verification all happen in your browser. The PDF is never uploaded.