Choose the language of the text, then drop one or more images on the box, or copy a screenshot and press Ctrl+V on this page. The text appears in the box below, ready to copy or save as a .txt file. Images are read one after another, and each one gets a heading with its file name when there is more than one.
What this tool reads well, and what it doesn't
The text recognition is Tesseract, an open source OCR engine first built at HP and later developed at Google, running in your browser through tesseract.js. It is good at printed text and weak at handwriting.
| Reads well | Struggles with |
|---|---|
| Screenshots of web pages, apps, chats and documents | Handwriting and signatures |
| Scanned letters, forms and book pages | Curved pages photographed close up, or text at a strong angle |
| Phone photos of printed paper in good light | Small, blurry or low contrast text, such as grey text on a photo |
| Receipts and invoices with clear print | Tables, where words come out in order but the columns are lost |
The number after the word count is Tesseract's own confidence score. Above about 90 percent the text usually needs only light checking. Below 70 percent the tool suggests checking closely, and the most common cause is the wrong language, a blurry image or text that is too small.
Languages
You can read English, Hindi, Malayalam, Tamil, Bengali, Arabic, Spanish, French, German, Portuguese, Indonesian and simplified Chinese. Pick up to three at once for documents that mix them, such as a form in Hindi and English. Only turn on the languages that are in the image: each extra one is slower and gives the engine more wrong guesses to choose from.
The first run downloads the reader, about 4 MB, and the data for each language you pick, between 0.7 and 3 MB. Your browser keeps them, so later runs start quickly.
Getting cleaner text
- Crop to the text before you drop the image. Photos of a whole desk with a page in the middle read worse than the page alone.
- Take screenshots at normal zoom. Small screenshots are enlarged automatically before reading, which helps, but a sharp original is better.
- For paper, hold the phone straight above the page in daylight, without flash glare.
- Join lines into paragraphs is on by default. It merges the line breaks inside each paragraph and rejoins words split with a hyphen at the end of a line. Turn it off to keep the line breaks of the original, for poems, addresses or lists.
Scanned PDFs
This tool reads images, not PDF files. For a scanned PDF, first turn its pages into images with PDF to JPG at 300 DPI, then drop those images here. If the PDF was made from a Word file and you can select its text, copy the text directly instead; OCR would only add mistakes.
Privacy
The reader runs in your browser tab, and your images are never uploaded. That matters for ID cards, bank letters and medical forms, which are the kind of document people most often need to copy text from.
After you copy the text
Text from OCR often carries stray line breaks and odd spacing. The AI Text Cleaner tidies spacing and invisible characters, the Case Converter fixes text that came out in capitals, and the Word Counter counts what you extracted.
