Image to Text (OCR)
Copy the text out of photos, screenshots and scanned PDFs, on your device.
Drop images or PDFs here
or select files. You can also paste a screenshot (Ctrl+V)
Files are processed on this device. Nothing is uploaded.
How to use Image to Text (OCR)
- Add one or more pictures or PDFs, or paste a screenshot with Ctrl+V (Cmd+V on a Mac).
- Choose the language the text is written in. Add a second language for mixed documents.
- Select Extract text, then correct anything in the box and copy it or download a .txt file.
How it works
This tool uses optical character recognition (OCR): it finds lines of text in a picture and recognises the characters. The engine is Tesseract, an open-source recogniser developed over many years (originally at HP, later sponsored by Google), running here as WebAssembly with its neural-network (LSTM) models.
The recognition engine and the language data are downloaded from this site to your browser the first time (about 6–8 MB for English), then the reading happens on your device. Your pictures, documents and the recognised text are not uploaded.
Small pictures are enlarged and very large ones reduced before reading, because the engine works best when letters are roughly 20–40 pixels tall. PDF pages are drawn at 200 DPI and read one by one. After each run you see the engine’s confidence (0–100%), its own estimate of how sure it is. Treat anything under about 65% with care.
Example
A sharp screenshot of a paragraph of English text is usually recognised without a single error. A phone photo of a printed receipt typically needs a few corrections, mostly in prices and codes.
Frequently asked questions
Is it really free and without limits?
Yes. There is no sign-up, no daily quota and no page limit other than 200 pages per run, because the work is done by your own device and costs us nothing.
Are my images uploaded?
No. The engine runs inside your browser. You can check by turning off your internet connection after the first run: it keeps working.
How can I get better results?
Use a sharp, well-lit, straight picture with dark text on a light background. Crop away everything except the text, choose the right language, and avoid photos taken at an angle. Scans at 300 DPI work best.
Can it read handwriting?
No. The models are trained on printed text. Neat block capitals sometimes work; cursive does not.
Can it extract text from a PDF?
Yes, including scanned PDFs: add the PDF and each page is read. If your PDF already has selectable text, the PDF to Text tool is faster and exact.
Which image formats are supported?
JPG, PNG, WebP, GIF, BMP, AVIF, TIFF and HEIC (iPhone photos), plus PDF.
Limitations
- Handwriting is not supported: the engine is trained on printed text.
- Ten languages are available: English, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi and Urdu. Chinese, Japanese and Korean are not included yet.
- Accuracy depends on the picture: blurry, skewed, low-contrast or very small text gives errors. Always check names and numbers.
- Tables and multi-column layouts come out as plain lines of text; the layout is not rebuilt.
Last reviewed November 9, 2026. How we build and test tools.