Pull the printed words out of any picture, from a photographed letter to a slide or a street sign, and copy them or save them as a .txt file. Nothing is uploaded, and there’s no account, watermark or limit.
Drop your pictures (PNG, JPG, WebP, HEIC, HEIF or AVIF) onto the drop zone, or tap “or tap to choose files”. You can add a whole batch at once.
Under “Text language”, check the language printed in the pictures: English, French or Arabic. Tick a second one only if a document really mixes two.
Keep “Enhance image” on, choose “Keep line breaks” or “Join into paragraphs”, then press “Extract text”.
Open “Show text” to read the result, then use “Copy text” or “Download .txt”. With several pictures, “Copy all text” and “Download all (.zip)” collect everything.
A photo of a page is only a grid of colored dots: your computer cannot search it, copy it or read it aloud. Optical character recognition (OCR) looks at those dots, finds the dark shapes that form lines, splits the lines into words and decides which letters the shapes most likely are. The output is ordinary text you can edit.
filetidy uses Tesseract, an open-source OCR engine built on an LSTM neural network, compiled to WebAssembly so it runs in a background worker inside your browser. The first time, the engine (about 4 MB) and a language model are downloaded from filetidy itself; your browser then keeps them, so later visits start straight away. Your pictures are never sent anywhere.
Every result gets a confidence rating based on how sure the engine was about each word. High means 85% or more and usually needs no correction; Medium (65 to 85%) means you should check names and numbers; Low means proofread the whole thing, often because the picture is blurry, tilted or handwritten.
This page accepts every common image type, including the HEIC and HEIF photos iPhones take and AVIF. For a specific job there are focused pages: JPG to text for phone photos of paper, screenshot to text for captures of apps and chats, PDF to text for documents, and dedicated Arabic OCR and French OCR pages.
iPhone & iPad
Live Text needs iOS 15 or later and works on one photo at a time.
Android
Windows
Retyping a page by hand takes minutes and introduces its own typos. On a sharp picture of printed text, OCR gets almost every character right in a few seconds, so your job shrinks to a quick proofread. It earns its keep most with long passages, lists of references or anything you would otherwise copy word by word.
Live Text on Apple devices and Google Lens on Android are excellent for grabbing a sentence from one photo. Where a browser converter helps is volume and files: drop forty pictures at once, get a .txt file per picture in a single zip, and do it on a laptop without an account or a particular phone.
OCR is not the right tool for handwriting, which Tesseract recognizes poorly, or for keeping a layout. Tables come out as lines of text without cells, and pages with several columns may be read straight across. If you need the formatting, plan to rebuild it after pasting.
Yes. There is no account, no watermark, no daily quota and no cap on the number of pictures. The recognition runs on your own processor, so filetidy has no per-page server cost to pass on.
Not reliably. The engine is trained on printed and typed text. Neat block capitals sometimes come out readable, but cursive notes usually produce a low confidence rating and many wrong words.
English, French and Arabic. The page starts with English selected; check another language under “Text language” before extracting. Picking the right model matters: the English one, for instance, strips French accents.
It is the average certainty the engine reported for the words it found. High is 85% and above, Medium is 65 to 85%, and Low is below 65%. A Low result is a hint to retake the photo in better light or to straighten it.
Pictures with two or more columns, captions next to photos, or text boxes scattered around a poster can be read line by line straight across the page. Crop the picture to one column at a time for a cleaner result.
No. The engine runs inside your browser tab and the pictures stay on your device. Only the engine and language files are downloaded, once, and they come from filetidy rather than a third-party service.
Last updated October 28, 2026
Enlarges small text, turns dark backgrounds light and evens out shadows. Try it off if letters come out broken.