Xargaar ToolsScans in. Text out.
Read text from photos, screenshots and scanned PDFs. Copy it, save it as a text file, or get a PDF you can search. The recognition runs in this page, so your documents are never uploaded.
How it works
Photos, screenshots or scanned PDFs. Several files at once is fine.
Cleanup is optional and may remove faint details or change the searchable PDF image. Compare the output with your original. Recognition uses one language per run. The free version supports 23 languages, including right-to-left and Asian scripts. Clean, printed synthetic samples were tested. That is not an accuracy benchmark for real documents. Scans, phone photos and handwriting can fail, so always check the result.
Edit the text in the box, copy it, download a .txt, or download a PDF with a hidden text layer so it becomes searchable. Edits in the text box affect .txt and copy only, not Word, CSV or the PDF layer.
Offline pack in preparation.
- The free page reads up to 5 pages per run. The planned pack removes that software cap, subject to browser memory.
- The planned download includes the engine and all 23 languages. Desktop Chrome offline operation has been tested; checkout is not yet available.
- Good for confidential scans, travel and locked-down networks.
The offline pack is not yet on sale.
Scanned statements
Scanned statement? Read it on the OCR page, download the table as CSV, then open that CSV in Statement Converter. Only synthetic statements were tested. Real noisy scans are less reliable. Check dates, columns, every amount and the balance check against the original before importing.
Questions
Are my documents uploaded?
No. The recognition engine is loaded into this page and runs on your device. There is no server that receives your files.
How accurate is it?
It works well on clean, straight, printed text at a good resolution. It struggles with handwriting, blur, curved pages, very small print and busy backgrounds. Read the result before you rely on it, especially numbers. Right-to-left searchable PDFs can copy letters or numbers in the wrong order. Check search and copy in your PDF viewer; use the reviewed text download when reusing the text.
Can it keep tables and give me a Word file?
The Word download inserts a page break between source pages and guesses tables from aligned columns. Text may flow onto extra pages. If a table is found you can download CSV for review in a spreadsheet. It does not copy fonts, colours, images or exact layout, and it can miss tables with no clear gaps between columns. Check the result against the original. Word and CSV use the original recognition, not your text edits. Word/table reading order is not verified for Hebrew or Arabic. This was tested on synthetic table images only.
Which languages?
The free tool reads 23 languages: English, French, German, Spanish, Italian, Portuguese, Dutch, Swedish, Polish, Czech, Romanian, Hungarian, Turkish, Greek, Russian, Ukrainian, Hebrew, Arabic, Hindi, Vietnamese, Indonesian, Japanese and Simplified Chinese. One language per run. Korean is not offered yet because we have not been able to test it.
What is a searchable PDF?
The pages are rendered from your scan and may change in quality or size. A hidden text layer is added behind the image so you can search and select text in a PDF viewer.
Why is the first run slow?
The page loads the recognition engine and the language data once (a few MB). Later runs can reuse those assets. Large scans take time because the work is done on your device.