Image to text in a few seconds
Upload a photo of a page, a screenshot or a scanned PDF and the text appears in the window. Copy it or download it as a TXT file. Recognition runs right in your browser.
How it works
- Upload an image
Drop a photo, screenshot or scanned PDF. JPG, PNG, WebP and iPhone HEIC all work.
- Choose the language
The page language is selected by default, together with English for mixed text. Pick a single language for better accuracy.
- Take the text
Copy the result with one click or download it as TXT. The first run takes longer: the browser downloads the models once, up to 10 MB.
What works well and what works worse
Works well
- Screenshots: almost no errors
- Photos of printed pages in even light
- Scanned contracts, invoices and letters
- Mixed-language text, such as German with English
- iPhone HEIC photos, no conversion needed
Works worse
- Handwriting: the engine is built for print
- Blurry shots, glare and strong tilt
- Tables: text comes line by line, without cells
- Small print on a busy background
- Formatting: bold, fonts and columns are not kept
The image never leaves your browser
- The file is read only inside your tab. Recognition is done by the Tesseract engine running in the browser.
- Only the engine and language models are downloaded from our site. On the first run that is up to 10 MB; after that they come from the browser cache.
- You can check this: open developer tools (F12, or Cmd+Option+I on a Mac), go to the Network tab and run recognition. There will be no request carrying your image.
- We see neither the file nor the extracted text. Analytics only receives anonymous events: “extracted”, the page count as a range, and the time taken.
Before you send important data such as numbers, amounts or names, check them against the original: OCR can confuse similar characters, like “0” and “O”.
Frequently asked questions
How accurate is the recognition?
Printed text on a sharp screenshot or a straight photo is recognised almost without errors. Accuracy drops with blur, small print, busy backgrounds and strong tilt. Check important numbers and names against the original.
Does it read handwriting?
Poorly. The engine is built for printed text, so handwriting turns into errors. Neat block letters are sometimes recognised, but do not count on it.
Can I extract text from a scanned PDF?
Yes. Upload a PDF of up to 20 pages and the converter reads every page. This helps with scans where the text cannot be selected. If the PDF is not a scan and you need an editable Word file, use the PDF to Word converter.
Which languages are supported?
English, German, French, Italian, Spanish and Russian, each on its own or together with English for mixed text. A single language gives the best accuracy.
Why is the first run slower?
On the first run the browser downloads the engine and language models, up to 10 MB. They stay in the cache, so next time recognition starts right away.
Is it safe and free?
Yes. The file never leaves your browser; we never receive or store it. The converter is free and needs no sign-up.