Copy the text out of any photo, scan or screenshot instead of retyping it, in English and 20 other languages.
How to convert an image to text
- Pick the language of the text (and a second one for mixed text).
- Choose one or more images, drop them on the box, or paste a screenshot with Ctrl+V. JPG, PNG, WebP, GIF, BMP, HEIC (iPhone), TIFF and scanned PDFs all work.
- The text appears under each image with a confidence score and the words worth checking. Fix anything in the box.
- Copy the text, download it as a .txt file, or save a searchable PDF. If a page is sideways, press ↻ to turn it and read it again.
How it works
The text is read by Tesseract, the open-source OCR engine, through tesseract.js 7.0.0 (Apache-2.0), with the fast LSTM language models from the Tesseract project's tessdata_fast set (Apache-2.0). Everything runs as WebAssembly in your browser. Images are turned the right way up from their camera data, small text is enlarged, slight tilts are straightened, and each image is then read line by line. The models are 0.3 to 2.1 MB each; English is 2.0 MB.
Tips for accurate OCR
- Use a sharp, well-lit photo taken straight on, and crop it to the text.
- Pick every language in the image; reading Hindi text with only English selected gives nonsense.
- Printed text works best. Handwriting, very stylised fonts and text on busy backgrounds often fail.
- Low confidence on a clear page usually means it's sideways or upside down: press ↻.
Need to fix the grammar of the text you got? Paste it into the grammar checker. A PDF whose text you can already select is faster and exact with PDF to Text. A password-protected PDF asks for its password here; it is used in your browser only.
Frequently asked questions
Is it free, and are my images uploaded?
It's free with no sign-up or daily limit. The images never leave your device: the OCR engine runs in your browser, so it also works for private documents. The first use downloads the engine (about 4 MB) and each language model you pick (0.3 to 2.1 MB) once.
Which languages can it read?
English, Arabic, Bengali, Chinese (Simplified), French, German, Gujarati, Hindi, Italian, Japanese, Kannada, Korean, Malayalam, Marathi, Portuguese, Punjabi (Gurmukhi), Russian, Spanish, Tamil, Telugu and Urdu. Pick a second language for mixed text, such as Hindi and English on one form.
How accurate is it?
On sharp, straight photos and scans of printed text it usually gets nearly every word right, and each result shows Tesseract's own confidence plus the words it was least sure of. Blurry, dark, curved or handwritten text gives more mistakes, so check names, numbers and amounts before you use them.
Can I convert several images or a scanned PDF?
Yes. Add up to 20 images or PDF pages at once; each page of a scanned PDF is read like a photo. Then copy or download all the text in one file, or save one searchable PDF. If a PDF's text can already be selected, PDF to Text is faster and exact.
What is a searchable PDF?
Your image with the recognised text laid invisibly over it, so you can search, select and copy the words in any PDF reader while it still looks like the original. It uses the text as read, not your edits in the box.