Extract PDF text locally from digital or scanned pages. Password-protected files require their password. Urdu/Arabic OCR, Quran mode, entity extraction, DOCX export. No upload.
Usually not. Extract Text (auto) requests the hybrid extraction path, which can use a digital text layer. Force OCR every page explicitly requests recognition instead.
The page passes the file and password to browser-side processing. OCR downloads its recognition engine and language data when you use it; those downloads do not upload the PDF.
Not necessarily. This page can reload previously extracted results from a browser cache. More settings includes Clear extraction cache (all PDFs). Keep this in mind on shared devices.
Yes. Each page’s text area is editable. TXT and other page-text exports use the current results. Copy all copies all page text, regardless of search or export range.
The page offers automatic extraction and a forced OCR option, with per-page results. You can inspect warnings, edit extracted text and export selected pages in several formats.
A recognition result is not a verified transcription. Check reading order, punctuation and figures before using the text. A language choice and another recognition attempt may help, but neither guarantees correct output.
If the reviewed result for page 1 is “Invoice total: 42”, select page 1 and click TXT. The file begins with a UTF-8 byte-order mark, then “--- Page 1 ---”, a newline, and “Invoice total: 42”. Export adds the heading. Review recognition first; a scan may not return this exact sentence.
If page 1 contains “Alpha” and page 2 contains “Beta”, an MD export containing both pages has “## Page 1”, a blank line, “Alpha”, then two newlines, “---”, two newlines, “## Page 2”, a blank line and “Beta”. Setting the export range to 2 leaves only the second page heading and Beta. The filename follows the uploaded PDF’s name with .md replacing .pdf.
The file-selection handler requires application/pdf. This page sets no explicit file-size or page-count cap; an unreadable file or browser memory shortage can still stop extraction. Password-protected documents require the correct password.
Parsing, recognition and export run locally. OCR engine and language downloads use cdn.jsdelivr.net; do not assume they are available offline or downloaded only once. Scan resolution, contrast, skew, font and chosen language affect recognition. Verify names, numbers and important passages against the original.
Results are page text, not a guarantee of original layout or reading order. Review automatic header/footer removal and any recognition warnings. Exports include only the selected result pages; search filtering does not change that selection.
Browse all free tools · Guides and tutorials · PDF tools · Developer tools · Text tools · SEO tools