Extract text from a photograph, a screenshot or a scanned PDF, and download it as a .txt file or as a searchable PDF: the original pages with the recognised words laid invisibly over them, so the document looks exactly as it did but can be searched, selected and copied. Recognition runs page by page and can be stopped part-way through if you only need the first few pages.
OCR usually means uploading a document to somebody else's server, which is a problem for anything confidential. Here the recognition engine is downloaded to your browser and the file never leaves your machine. It is free to use, with no account and no upload limit, and it runs entirely in your browser: nothing is uploaded, and while the page stays open, a feature you have already used keeps working offline.
- Is my scanned document uploaded for recognition?
- No, and this is the tool where that matters most. The Tesseract engine is downloaded to your browser as WebAssembly and runs there, so a scanned passport, contract or medical record is never sent anywhere.
- Why is the first run slow?
- The engine and the trained data for your chosen language are downloaded on first use, which is a one-off cost your browser then caches. Recognition itself is your own processor doing the work, so a long document takes real time.
- The text came out as gibberish. What went wrong?
- Usually the wrong language, or a source image that is too low-resolution, skewed or badly lit. Pick the right language and press Extract text again - every page is read afresh in the new language. OCR is far more sensitive to input quality than people expect, and a sharper scan beats any setting.
- Can it read handwriting?
- No. Tesseract is trained on printed text and will produce nonsense from handwriting.
- Does it keep the layout of the original?
- The extracted text does not: you get the words, not the columns, tables or formatting around them. The searchable PDF does, because it is the original page itself with the text laid invisibly on top.
- How does it make a scanned PDF searchable?
- Each page of the PDF you dropped in is kept exactly as it was - not re-rendered or recompressed, so the file stays close to its original size - and Tesseract's recognised words are added as an invisible layer, each one placed over the word it was read from. A PDF reader can then find, select and copy the text, and because the words are stored as Unicode rather than in a Western-only font, non-Latin text such as Russian can be searched too. When every page comes from one PDF, the rest of that file, such as its bookmarks and links, is kept too. A dropped picture becomes a page of its own, and a page you have not read yet goes in with no text on it.