Image to text

What this tool does

Reading the letters in a picture and turning them into text is called OCR, optical character recognition. Pick a picture, drop it on the box or paste a screenshot with Ctrl+V, and the English (and Korean) in it appears in a text box you can edit. Several pictures are read one after another in the order you add them, and you can copy all of them at once or save them as one plain-text file.

The reader is Tesseract, the open-source OCR engine. Many OCR sites send your picture to their server; this page downloads the reader and its language data from this site once, and reads the picture inside your browser. That makes it a reasonable choice for pictures you would rather not hand to a stranger: contracts, receipts, letters, forms with personal details.

Words the reader is not sure about are framed on the picture, so you know where to look before you use the text. Light text on a dark background is switched to dark-on-light before reading, small text in screenshots is enlarged, and uneven light in a photo is evened out first.

How to use it

1. Press 'Choose pictures' and pick a photo or screenshot, or drop it on the box. For text on your screen, take a screenshot (Windows: Win+Shift+S, Mac: Cmd+Control+Shift+4) and press Ctrl+V (Cmd+V on a Mac) on this page.
2. The first time, the reader and its language data come from this site (about 5.9 MB for English, 7.5 MB for English + Korean). A bar shows the download; your browser usually keeps the files, so later visits start sooner.
3. The text appears in the box next to the picture. Fix any wrong letter right there, and check the words framed in coral on the picture.
4. Press 'Copy' and paste into a note, a document or an email. With several pictures, use 'Copy all' or 'Save as text (.txt)'.

Which pictures read well, and which do not

Screenshots, scanned pages and straight photos of printed text read best. On the five test pictures made for this page — an English notice, a Korean notice, mixed Korean and English, a dark-mode screenshot with small text, and a tilted, unevenly lit phone photo — more than 95% of the letters came out right on every one.

Handwriting, badly blurred photos, tiny text, decorative fonts and vertical writing do not read well. Services that read on large servers, such as Google Lens, may do better with those — but your picture then goes to that company. Tables come out line by line, without columns.

Joining the lines back into paragraphs

Text in a picture is broken into lines to fit the page. Pasted as it is, lines break in the middle of sentences. Turn on 'Join lines' and the lines of each paragraph become one line again; a word split with a hyphen at the end of a line (exam-ple) is put back together.

Leave it off for poems, addresses and lists, where the line breaks mean something. The text box changes as soon as you switch it, and text you have edited by hand is left alone.

Korean, and Korean mixed with English

With 'English + Korean' both sets of letter data are used together, so a Korean notice with file names, email addresses or English words in it is read in one go. Korean needs one more step: Tesseract often puts a space between the syllables of one Korean word, so this page measures how much empty space there really is between the letters in the picture and keeps a space only where the words were spaced.

Changing the language reads the pictures you already added again; text you have edited by hand is kept.

Your data

Pictures you choose or paste are opened and read only inside this browser. The pictures and the text read from them are never sent anywhere or stored, so they are gone when you close the page. The first time, the reader and the data for the languages you chose are downloaded from this site; your browser keeps them, so later visits usually do not download them again. The only settings remembered on this device are your language choice and 'Join lines'.

Questions and answers