Image to text (OCR)
Extract text from photos, screenshots and scans in your browser, then edit it, copy it or save it as a .txt file. Nothing is uploaded.
How to use
Choose one or more images, drag them onto the tool or paste a screenshot with Ctrl+V. The tool reads them one after the other and puts all the text in one box, where you can correct it before you copy it or download it as a .txt file. With several images, each one starts with a line showing its file name, so you know where every part came from.
Choose the language the text is written in before you add the images. If you change it later, or change the reading order, the images are read again. The language matters because it tells the engine which letters and accents to expect: Portuguese text read as English can lose accents such as ç and ã. The reading order matters for layout: line by line keeps each row of a table or receipt together, while columns and blocks reads a page with columns one column at a time.
The recognition engine is Tesseract, the open-source OCR engine, running in your browser as WebAssembly. Your images are not uploaded to nTools or to any other service. The first time, the browser downloads the engine (about 1.5 MB) and the data for the language you chose (0.7 to 3 MB); after that they normally come from the browser's cache.
The engine was trained on printed text. Handwriting is not supported, and blurred photos, text at an angle, very small letters, decorative fonts and light text on a busy background all produce errors. Small images, such as most screenshots, are enlarged automatically before they are read, which helps with small interface text. The confidence shown is the engine's own estimate, averaged over the text it found: a useful warning sign, not a guarantee, so check names, numbers and amounts against the image.
TIFF files are read page by page, including multi-page scans and ZIP-, LZW-, JPEG- or fax-compressed files. JPEG and HEIC photos are turned the right way up using their orientation information, but text that is sideways or upside down in the image itself is not rotated. Each image can be up to 30 MB, 40 megapixels and 12,000 pixels per side, and up to 20 files can be read at a time (a multi-page TIFF counts as one).
Example
Screenshot of an error message (PNG)The message as text you can paste into a search or a support ticket.Photo of a printed letter, taken straight on (JPEG)The text of the letter, ready to correct and save as a .txt file.Multi-page TIFF from a scannerThe text of every page, each one headed with the file name and the page number.Photo of a receipt, read line by lineEach item next to its price on the same line. Check the amounts: a 5 can be read as an S, or a comma as a full stop.
Frequently asked questions
Are my images uploaded?
No. The text is recognised in your browser, and the images and the text stay on your device. The browser downloads from nTools only what it needs to read them: the recognition engine, the language data and, for TIFF or HEIC files, a decoder. No image or text is sent with these requests.
Can it read handwriting?
No. The engine was trained on printed text, so handwriting gives poor results unless it is very regular, close to block capitals.
Why is some text wrong or missing?
The usual causes are a blurred or tilted photo, small or faint letters, a busy background, or the wrong language selected. Take the photo straight on and in good light, or crop the image to the text, and try again. For tables and receipts, use the line-by-line reading order.
Can I extract text from a PDF?
Not here: this tool reads images. If the PDF was scanned, convert its pages to PNG images at 144 DPI with the PDF to images tool and read those. If you can select the text in the PDF, it is already text, so copying it is more accurate than OCR.
Does it keep the formatting?
No. The result is plain text with line breaks, without fonts, bold, columns or tables. That is what a .txt file holds.