SweetcraftOCR PDF
Recognize text in scanned PDFs and images with OCR, and rebuild them as searchable PDFs with an invisible text layer. Runs in your browser; files are never uploaded.
About this tool
A scanned PDF is a photograph of a document: it looks complete, but there is no text inside to search, select or copy — which is why extractors return empty files for scans. This tool runs optical character recognition (Tesseract) on every page that has no usable text layer, at about 300 DPI after a grayscale and contrast cleanup, and gives the words back.
You get two files: the recognized text as a plain .txt, and a searchable PDF that keeps the scanned pages looking exactly as they were while carrying an invisible text layer where the words were found, so Ctrl+F and text selection work in any reader. Pages that already had real text are copied unchanged. Recognition runs entirely in your browser — the first run downloads the OCR engine and language data from a public CDN, and your documents are never uploaded.
How to use OCR PDF
- Choose a PDF or an image (PNG, JPG, WebP) — up to 20 MB; PDFs up to 50 pages.
- Recognition starts automatically: pages with real embedded text are kept as-is, scanned pages are OCRed.
- Download the recognized text (.txt) and the searchable PDF.
Keep the original files and inspect the exported result before publishing or sharing it.
Frequently asked questions
Which languages can it recognize?
English, Simplified Chinese, or both together — pick before you choose the file, or change it and run again. Clean printed text recognizes best; handwriting is not supported.
What is a searchable PDF?
The output looks like your scan, but carries an invisible text layer placed where the words were found, so you can search, select and copy the text in any PDF reader.
Are my files uploaded for OCR?
No. Recognition runs in this tab. The first run downloads the OCR engine and language data (a few MB, cached by your browser afterwards) — that download is the only network traffic; your documents never leave the device.
My PDF already has text on some pages — what happens?
Pages with a real text layer are copied into the output unchanged; only pages without usable text are recognized. The result is one searchable PDF either way.
How accurate is it?
On clean printed pages (200 DPI or better) expect near-complete word recovery. Skewed, faint or photographed pages recognize worse — straighten and re-scan when the words matter.
Why a 50 page limit?
OCR is heavy work for a browser tab. Split longer documents with Split PDF and OCR the parts.