Image to Text Converter — Free OCR Online, In Your Browser
Extract text from screenshots, scans and photos with OCR that runs on your device. English, Chinese and 14 more languages; copy it or save a .txt.
First use downloads about 3 MB of language data, then it is cached. Picking the right language matters more than anything else for accuracy.
Works best on screenshots and clean scans of printed text. Handwriting, stylised fonts, photos taken at an angle and very small or blurry text give poor results.
Text recognition runs on your device with Tesseract.js; your images are never uploaded. Only the OCR engine and language data are downloaded. Browse all free tools.
Found a problem, or something that could be better? We read every message and we fix things quickly.
This free image to text converter pulls the words out of a screenshot, a scanned page or a photo of a document using OCR (optical character recognition), and it does the recognition in your browser, so the image is never uploaded. Drop in a PNG, JPG, WebP, BMP or GIF, or just press Ctrl+V to paste a screenshot, choose the language, and the recognised text appears in an editable box you can copy or save as a .txt file. Several images can be queued and are read one after another, and the results saved as one combined text file or as a ZIP of separate .txt files, one per image.
The engine is Tesseract, the open-source OCR engine behind many desktop and server tools, compiled to WebAssembly. On clean printed text, such as screenshots, PDFs saved as images, book pages and receipts photographed flat, it is accurate enough that you mostly fix the odd character. It is not good at handwriting, decorative fonts, text on busy backgrounds, or small blurry photos, and no free OCR is; the sections below explain how to get the best result and when another approach is better.
How to use
- Choose the language of the text first. For a page that mixes English and Chinese, pick one of the English + Chinese options.
- Drop one or more images on the box, click to choose files, or copy a screenshot and press Ctrl+V (⌘V on a Mac) anywhere on the page.
- Wait for recognition. The first run downloads the OCR engine and the language data (a few MB, cached afterwards); later runs start straight away.
- Check the text in the box and correct anything it misread. The confidence figure gives a rough idea of how much checking is needed.
- Copy the text or download it as a .txt file. With several images, Copy all and Download all as one .txt combine them under each file name, and Download separate .txt files gives a ZIP with one file per image.
Frequently asked questions
How do I extract text from an image for free?
Open this page, choose the language, and drop the image on the box or paste it with Ctrl+V. The text appears a few seconds later in an editable box with Copy and Download .txt buttons. There is no account, no limit on the number of images and no watermark, because the work happens on your own device rather than on a server.
Can I extract text from multiple images at once?
Yes. Select, drop or paste as many images as you like; they are queued and recognised one after another, each with its own progress, editable result and error message if it fails. When they are done you can copy all the text, download it as one combined .txt with a header per image, or download a ZIP containing a separate .txt file for each image, named after it.
Are my images uploaded to a server?
No. Tesseract runs inside your browser as WebAssembly. The only downloads are the OCR engine (from this site) and the trained data for the language you pick (from the jsDelivr CDN), both cached after the first use. Your image and the recognised text stay on your device, which makes the tool suitable for invoices, IDs and internal documents.
Can it read Chinese text?
Yes. Choose Chinese (Simplified) or Chinese (Traditional), or an English + Chinese option for mixed pages such as menus, slides and technical screenshots. The tool removes the spaces Tesseract tends to put between Chinese characters. Accuracy on clear printed Chinese is good; small text, vertical layouts and calligraphic fonts are much weaker.
Can it read handwriting?
Only very neat, print-style handwriting, and even then expect many errors. Tesseract is trained on printed and typeset text. Cursive notes, signatures and whiteboard photos generally come out as noise. For handwriting, the OCR in Google Lens, Apple Live Text or Microsoft OneNote uses models trained on handwriting and does much better.
Why is the recognised text full of mistakes?
Usually one of four things: the wrong language is selected, the text is too small or blurry (aim for letters at least 20 pixels tall), the photo is taken at an angle or in poor light, or the text sits on a patterned or coloured background. Crop to the text, retake the photo straight on in good light, or zoom in before taking a screenshot, then run it again.
Does it keep the layout, tables or formatting?
No. You get plain text with line breaks and blank lines between blocks. Columns are usually read one after the other, but tables become lines of words separated by spaces, and bold, fonts and images are not preserved. For a document you need to edit with its formatting, OCR it in Word, Google Docs or Acrobat instead.
Getting accurate results from OCR
OCR accuracy depends far more on the image than on the engine. A few habits make the difference between a clean result and a page of garbage:
• Pick the right language. Tesseract uses a language model to choose between similar shapes; English data will turn French accents and Chinese characters into nonsense.
• Size matters. Text should be at least 20 pixels tall in the image. The tool already doubles the size of small images, but it cannot invent detail that a blurry photo does not have.
• Crop to the text. Remove toolbars, photos and decorative borders first (the image cropper works), so the engine does not try to read them.
• Shoot straight on, in even light. Perspective, shadows and curved book pages are the most common causes of errors in phone photos.
• Dark text on a light, plain background reads best. Light text on dark backgrounds usually works; text over photos or patterns often does not.
Screenshot to text: the fastest workflow
For text you can see on screen but cannot select, such as an error dialog, a video frame, a locked PDF or text inside an image on a website, a screenshot plus paste is quickest. On Windows press Win+Shift+S, on a Mac Cmd+Ctrl+Shift+4 (which copies to the clipboard), select the area, then come here and press Ctrl+V or ⌘V. The screenshot is recognised immediately with the selected language.
Screenshots are the best possible OCR input: perfectly sharp, straight and evenly lit. If the text is small on screen, zoom the page in (Ctrl and +) before taking the screenshot; the extra pixels improve accuracy noticeably.
Browser OCR versus cloud OCR services
Cloud OCR from Google, Microsoft and Amazon uses larger models and handles handwriting, complex layouts and poor photos better than Tesseract. The cost is that your image is uploaded to their servers, usually with an account and per-page pricing once you go past a free tier. Phone features such as Google Lens and Apple Live Text are excellent and free, but work on the device they are on.
This tool sits at the other end: nothing leaves your computer, there is no account or quota, and it works on any desktop browser. For printed text in screenshots and scans that is usually all you need. If you have scanned PDFs rather than images, export the pages as images first (the PDF converter does this); if your PDF already contains selectable text, PDF to Word extracts it directly without OCR.
Related free tools
- PDF to Word — Turn a text PDF into an editable Word document: paragraphs and headings extracted in your browser. Text only, and honest about it.
- Text Cleaner — Remove line breaks, extra spaces, duplicates and invisible characters, sort lines, and find and replace, in one pass.
- Word Counter — Count words, characters, sentences and reading time, and check X, LinkedIn, Instagram, YouTube, SEO and SMS limits as you type.
- Image Cropper — Crop images to any size or aspect ratio in your browser — square, 16:9, 4:3, or freeform.