OCR — Scan to Text

Drop a scan or photo here

or click to browse — PDF, JPG, PNG, WebP, up to 25 MB

Scanned & image-based PDFs up to 25 pages — password-protected PDFs not supported

What OCR does

Reads scanned PDFs with no text layer
Transcribes photos of documents
Preserves headings, lists & tables
Marks unreadable regions as [illegible]
Copy or download as .txt / .md
Up to 25 pages per document

ℹ️ For best results with photos, keep the document flat and well-lit. Already-selectable text? PDF to Word is faster.

When you need OCR (and when you don't)

Most PDFs created from Word, Google Docs, or a print-to-PDF already carry a text layer — the words are stored as selectable text. For those, PDF to Word and Chat with PDF work directly and instantly, with no daily limit.

OCR is for the other kind: a scan from a copier, a fax, or a photo taken on your phone. There, the page is just an image — selecting text gives you nothing. AFileFix's OCR reads the pixels with AI and reconstructs the text, so you can copy, search, and reuse it.

How AFileFix OCR works

Upload a scanned PDF or a photo. The document is sent to Anthropic's Claude API, which reads each page as an image and transcribes the text verbatim — preserving headings, lists, and tables as Markdown. It does not summarize or rewrite; anything genuinely unreadable is marked [illegible] rather than guessed.

Longer PDFs are processed a few pages at a time, in order, which keeps each job fast and reliable. When it's done, copy the text or download it as a .txt or .md file. Your document is held in memory only, forwarded to Anthropic for processing, and never written to our disk.

Why the 25-page limit?

Reading a page as an image is much heavier than reading a text layer, so OCR is capped at 25 pages per document to keep it fast and free. If your file is longer, use Split PDF to break it into 25-page parts and OCR each part — both tools are free and run back to back.

What works best

Best results: clean scans and sharp, well-lit photos of printed text — letters, forms, invoices, book pages, contracts. Tables and multi-column layouts are preserved.

Modest results: faint or low-contrast scans, skewed phone photos, dense handwriting, and very small print. The AI still transcribes, but expect some [illegible] markers — rescan or rephotograph for a cleaner pass.

Whatever the input, treat OCR output as a strong first draft: proofread before relying on it for anything important.

Frequently asked questions

How is this different from PDF to Word or text extraction?

PDF to Word and Chat with PDF read the text layer already inside a PDF — they don't work on scanned documents, where the page is just an image of text. OCR reads the pixels with AI, so it can transcribe scanned PDFs and photos that have no selectable text at all. If your PDF already has selectable text, use PDF to Word or Chat with PDF instead — they're faster and free of the daily OCR limit.

Why is there a 25-page limit?

OCR sends each page to the AI as an image, which is slower and more costly than reading a text layer. The 25-page cap keeps each job fast and keeps the tool free. For longer documents, use Split PDF to break the file into 25-page parts, then run OCR on each part.

How accurate is the transcription?

Very good on clean scans and clear photos — printed text, forms, and tables transcribe faithfully, and layout is preserved as Markdown. Accuracy drops on faint scans, heavy handwriting, skewed photos, or low-resolution images. Anything the AI genuinely can't read is marked [illegible] rather than guessed. Always proofread the output before relying on it.

What can I upload?

Scanned or image-based PDFs (up to 25 pages) and photos or screenshots in JPG, PNG, or WebP. Files must be under 25 MB. Password-protected PDFs aren't supported — unlock them first.

For AI-processing details and Anthropic's API retention terms, see our Privacy page.

Extract text from a scan now

Free, no signup. Turn a scanned PDF or photo into copy-ready text.

AFileFix

PDF tools + AI for everyone

© 2026 AFileFix. All rights reserved.

Made with care for privacy