Text-Rex

Grab the text out of pictures. Right-click any image, drag a box over anything on screen, or read every image at once.

Turn any picture into text you can actually use. Text-Rex reads the words inside images on any web page, so you can copy, edit, search or save them - screenshots, photos of documents, memes, scanned pages, charts, error messages, anything with writing in it. THREE WAYS TO USE IT - Right-click any image and choose "Grab text from this image". This reads the ORIGINAL image file at full resolution, so it catches small print that a screenshot would blur away. - "Drag a box over the page" for anything else you can see: text inside a video frame, a canvas, a map, a PDF in the browser viewer, or a site that blocks right-clicking. - "Every image on this page" reads a whole gallery in one go. Results appear in the side panel, where you can edit them before copying or saving as a .txt file. EVERYTHING HAPPENS ON YOUR COMPUTER There is no server, no account and no upload. The text reader and its language data are built into the extension, so your images never leave your machine - and it keeps working with no internet connection at all. PLEASE READ - HOW ACCURATE IS IT? Text-Rex reads text using Tesseract, the long-established open-source recognition engine, running entirely inside the extension on your own computer. Tesseract is very good, but no text recognition is perfect, and you should expect occasional mistakes. They are most likely when: - the background behind the text is busy, dark, or low in contrast - the text itself is blurred, very small, handwritten, or in an unusual or highly decorative typeface - pictures, borders, patterns or specks sit close to the text, which the engine can misread as letters This is exactly why the text box is fully editable. Read the result over and correct anything it got wrong before you copy or save it - usually a couple of characters at most, and far faster than typing the whole thing yourself. Old scans, faded print and photographs taken at an angle are where you will see the most errors. Clean screenshots and clear printed pages are usually near perfect. IF A WHOLE IMAGE READS BADLY, TRY A SMALLER PIECE OF IT This is the single most useful trick with this extension. Reading an entire image at once asks the engine to cope with everything in it simultaneously - several columns, headings at different sizes, photographs, captions, borders, the edges of the page. Any one of those can throw it off, and a single troublesome patch can scramble the text around it too. Selecting only the part you actually want usually fixes it. Press "Drag a box over the page" and draw around one paragraph, one headline, or one column. With nothing else competing for its attention the engine reads that block far more reliably. Small print in particular often comes out better this way, because a small selection is enlarged before it is read. So if a full-image grab comes back messy, do not give up on the image - take it a piece at a time. Tick "add to what's there" and each piece is added to what you have already collected instead of replacing it, so you can work across a page, or across several pages, and end up with everything in one place. LANGUAGES Twelve are included: English, Dutch, French, German, Greek, Italian, Polish, Portuguese, Russian, Spanish, Turkish and Ukrainian. These are not an arbitrary selection, and the reason is the same engine described above. Tesseract performs genuinely well on Latin, Cyrillic and Greek scripts, so those are what is offered. Languages written in other scripts, such as Chinese, Japanese, Korean and Arabic, are deliberately left out rather than included and disappointing you - accuracy on them is not yet good enough to be worth putting in front of you. As Tesseract improves, more languages will be added. There is also a "plus English" option, because real pages constantly mix English words into other languages - brand names, menu labels, technical terms. CREDITS Text-Rex is built on Tesseract, the open-source text recognition engine, used under the Apache License 2.0. A full copy of the license and the attribution notices is included inside the extension.