Image to text
Copy the text out of a photo, screenshot, scan or PDF. It is read on your own device; the file never leaves it.
Add a photo, screenshot, scan or PDF. The first OCR downloads the reader (about 6 MB, once) from this site.
word readless than 60 % sure: check itTap a word to find it in the text.
Text
Copy takes the image or PDF page shown. The .txt holds everything you added, each image under its file name and each PDF page under its own heading (with read from text or read by OCR). Edits you make in the box are kept in both.
Your images, PDFs and their text stay on this device. They are read inside this page by Tesseract OCR and PDF.js, both served from this site. No upload, no account; once the reader has loaded it keeps working in airplane mode.
How to copy text from an image
- Add the image or PDF. Tap Choose images or a PDF, drop files on the box, press Ctrl+V (⌘V on a Mac) after copying a screenshot, or tap Camera on a phone to photograph a page. Add several and they are read one after another. For a PDF, type the pages you want in PDF pages first (1-3, 7); left empty, every page is read.
- Pick the language if the text is not English. German, French, Spanish, Italian, Portuguese, Dutch and Danish each download once, between 0.6 and 2.8 MB; the size is in the list before you pick.
- Fix the picture if the text comes out wrong. Rotate a sideways photo, crop to the part you want (only the box is read), or tick Boost contrast for a faded receipt or a grey scan. Each change reads the image again.
- Check the shaded words. Every word found is outlined on the image; shaded ones are the reader's guesses. Tap one to jump to it in the text and fix it.
- Copy or save. Copy text takes the image you are looking at; Download .txt saves all of them in one file, each under its file name.
What it cannot do
- Handwriting is read poorly. The models are trained on printed text; neat block capitals sometimes work, joined-up writing mostly does not.
- Table layout becomes plain lines. Columns come out as words in reading order, one row per line, without cell borders or alignment.
- The first read downloads about 6 MB once. That is the reader (3.9 MB) plus the English data (1.9 MB), from this site; other languages add 0.6 to 2.8 MB each. On a slow phone connection the first image takes a while.
- No searchable-PDF output. You get the text (copy or .txt), not a new PDF with a hidden text layer over the scan.
- Password-protected PDFs cannot be opened. The tool says so instead of guessing; remove the password in the app that made the file, or print it to a new PDF, and add that copy.
- Up to 50 PDF pages at a time. Each page is kept as a picture so you can rotate or crop it, which costs memory on a phone. For a longer PDF, read it in ranges such as 1-50, then 51-100.
- A page counts as scanned when its text layer has fewer than 20 letters or digits. A page that mixes typed text with a pasted-in picture of text is read from its text only; tap Read this page by OCR to read the picture too.
What reads well, and what does not
| Image | Typical result |
|---|---|
| Screenshot of a chat, web page or app | Near-perfect: sharp, high-contrast text |
| Flat scan of a printed page | Very good when the print is sharp; a low-resolution scan of small print loses accents and punctuation first |
| Phone photo of a document | Good if it is flat, in focus and well lit; curved pages and shadows cost words |
| Receipt on thermal paper | Fair; faded ink improves with Boost contrast |
| Text on a busy photo (signs, labels) | Patchy; crop tightly to the text |
| PDF saved from Word, a web page or an e-mail | Exact: read from the text the PDF already carries, no OCR and no download of the reader |
| Scanned PDF (from a copier or a scanner app) | As good as a flat scan; each page is drawn at 2× (144 dpi) before reading, so 10-point print is about 20 pixels high |
Images are enlarged up to 2× before reading, because small screenshot text (11 to 13 pixels high) reads badly at its own size. Very large photos are scaled down to 4,000 pixels on the long side, which keeps a 12-megapixel photo fast on a phone without losing normal-size print.
Questions
How do I convert a JPG to text?
Tap Choose images (or drop the JPG on the box), wait a few seconds while it is read, then edit the text in the box and tap Copy text or Download .txt. PNG, WebP, BMP and GIF work the same way; for an animated GIF the first frame is read.
How do I get text from an image on an iPhone?
Open this page in Safari. Tap Camera to photograph a page, or Choose images to pick a photo or screenshot from Photos or Files. If you copied a screenshot, tap Paste. Then tap Copy text and paste it into Notes, Mail or Messages.
Is it free?
Yes. No account, no email, no limit on images and no paid version. The reading happens on your device, so there is no server bill per image.
Is my image uploaded?
No. The image is decoded and read inside this page by Tesseract, an open-source OCR engine compiled to WebAssembly and served from this site. The page downloads the engine and the language data and sends an anonymous page-view count; none of those requests contains your image, its file name or its text. Once the reader has loaded you can switch to airplane mode and keep going.
How do I OCR a PDF?
Drop the PDF on the box. To read only some pages, type them in PDF pages first, for example 1-3, 7 (or 5- for page 5 to the end). Each page is checked on its own: a page that already carries text, such as a PDF saved from Word, is read from that text exactly and labelled read from text; a scanned page has no text layer, so it is drawn at 2× and read by OCR, labelled read by OCR. Progress shows the page being read, and Cancel stops the rest. Download .txt saves every page under its own heading. The PDF is opened inside this page and never uploaded; a password-protected PDF cannot be opened.
Why are some words shaded?
Shaded words are ones the reader is less than 60 % sure of. Check those against the image: tap a word on the image and it is selected in the text, so you can fix it in place.