Image & PDF to Text (OCR)
PopularTurn a scan, a photo of a page, a screenshot or a scanned PDF into text you can edit and copy, or into a searchable PDF. Korean and English, read on your device by an open-source engine; doubtful words are marked so you know what to check.
Runs on your device. Nothing is uploaded.
Language
Korean and English together reads mixed pages; one language alone is a little faster and a little surer.
Reading
The reader (about 6 MB) is fetched on first use and kept by the browser. Pages are read on this device.
No file at hand?
How to use the Image & PDF to Text (OCR)
- Drop scans, photos of documents, screenshots or scanned PDFs on the page, paste a picture with Ctrl+V, or click to choose. Each PDF page becomes one page in the list. Nothing is uploaded; the reader runs in your browser.
- Pick the language: Korean and English together for mixed pages, or one of them alone. Pick the layout: Page for a document in one column (letters, notices, forms), Columns for newspaper-style pages, Block for a screenshot of one passage, Line for a single line such as a label.
- Reading starts by itself and shows its progress page by page. Click a page to see it with the doubtful words outlined in red (Words outlines everything that was read) next to its text, which you can edit.
- Copy the text of a page or of everything, save a .txt, or save a searchable PDF: the pages exactly as they were, with the recognised words placed invisibly underneath so the file can be searched and its text selected. Send the result on to the Text Cleaner or the PDF Toolkit.
What it does well
It tells you what to check. Every word comes with a confidence, and the doubtful ones are outlined on the page beside an editable text box. A reader that only prints text leaves you proof-reading everything; this one points at the few words that need a look.
It gives you a real searchable PDF. The pages are kept exactly as they were (a PDF's own pages are copied, not re-rendered) and the words are placed invisibly where they were read, with an embedded font for Korean, so any reader can search, select and read the file aloud.
Nothing leaves the device. Documents are the files people least want to upload: contracts, notices, forms with names on them. The engine, the models and the fonts are served from this site and run in your browser, which is also why there is no queue and no page limit beyond your patience.
Frequently asked questions 8
What kind of pictures does it read well?
Printed text: scans, photos of printed pages taken straight on, screenshots, and PDFs that are only pictures of pages. The sample on this page (an English library notice at scan resolution) came back without a single wrong character in the Page layout; a Korean and English notice typeset the same way reads at 97%, the errors being a few wrong syllables in bold headings and a stray digit. It reads poorly from handwriting, from photos taken at an angle or in bad light, and from very small or decorative type; for a street sign or a menu, a phone's own camera text feature does better.
Does it handle Korean and English on the same page?
Yes. The default reads both at once, so a Korean notice with English terms, numbers and units comes out as one text. Keep both on even for a page that looks all Korean: dates, codes and units are Latin, and on the Korean and English notice we measure with, the Korean model alone read 76% against 97% for the pair. English alone is right for an English page, though the pair read the English sample on this page just as completely.
What is a searchable PDF?
A PDF that still looks exactly like the scan, with the recognised text placed invisibly over each word. Readers can then search it, select and copy its text, and screen readers can read it. For a PDF input the original pages are kept as they are; pictures become pages of the same size. Korean text in the layer uses an embedded font, so it survives in any reader.
Why are some words outlined in red?
The reader gives each word a confidence. Words under 60% are outlined so you know where to look: a smudge, a stamp over the text, an unusual font. Edit them in the text box before copying or saving; the searchable PDF carries what was read, not the edits.
How long does it take, and what does it download?
The first use fetches the reader (about 4 MB) and the language models (Korean 1.6 MB, English 4 MB), which the browser then keeps. A scanned A4 page takes a few seconds; a photo from a phone about the same once it has been scaled. Pages are read one after another with a progress bar, and you can stop at any time.
My photo came out as nonsense. What helps?
Straighten and crop it first (the Image Resizer can crop), make sure the text is dark on a light background, and that the characters are at least about 20 pixels tall; the Enlarge option does that for screenshots. Crooked scans and pages lying on their side are stood up before they are read (Stand tilted and sideways pages up, on by default): the Korean and English notice we measure with read 59% tilted by 4° and 98% once straightened, 0% lying on its side and 97% stood up. A page upside down is only turned when its first read comes out poor.
Is anything stored or sent?
No picture, page or text leaves your device. The history keeps only how many pages and words were read. The reader is Tesseract, an open-source engine, served from this site and run in a Web Worker in your browser.
Can it read tables, columns or vertical text?
Columns and tables are read in reading order with the Page layout. Keep table columns (on by default) separates the cells of a table with tabs, so the rows paste into a spreadsheet as columns: on the schedule we measure with (5 columns, 6 rows with the header) every row split into the right cells and 29 of 31 cells were read exactly. A tilted table splits less reliably than a straight one, even once straightened. Vertical Korean text and handwriting are outside what this reader does well.
Rate this tool
One tap is enough; a note is optional. Only the operator reads them.
More free tools
- PDF Toolkit Merge, split, reorder, rotate and compress PDFs, add page numbers or a watermark, and convert images, Word and Excel. Everything runs on your device.
- Sign PDF & Documents Draw, type or upload your signature and place it with initials, dates, text and check marks on PDFs or photos. PDF text stays selectable.
- Label Maker Design and print address, shipping and product labels on Avery-compatible Letter and A4 sheets. Mail merge from Excel, barcodes, true-size PDF.