How to use it
- 1 Add an image. Drag a JPG, PNG, WebP or HEIC image onto the box, or tap choose an image. Images up to 10 MB work, one at a time.
- 2 Pick the text language. Choose English, Spanish, French, German, Italian or Portuguese from the list, to match the text in the image.
- 3 Extract the text. Tap Extract text. The first run downloads the recognition model, which takes a moment; later runs start faster.
- 4 Copy or save. Check the text against the image, fix anything misread, then tap Copy text or save it as a .txt file.
Use this tool to extract text from an image when retyping it would take longer: a page of a printed handout, a screenshot of an error message, a menu, a label, a slide someone shared in a chat. You get plain text you can paste anywhere.
How it works
The tool uses Tesseract, a long-running open-source OCR engine, compiled to run inside a web page. When you tap Extract text, it loads the engine and the model for the language you picked, reads the image line by line and puts the result in the text box.
The first run downloads that model, a few MB. Your browser keeps it, so later runs start faster. When it’s done, the page shows how many words it found and a confidence score, which is the engine’s own estimate of how sure it was.
HEIC photos from an iPhone are converted to JPG in the browser first, so you can drop them in as they are.
Privacy: the image stays on your device
The recognition happens on your computer or phone, inside the browser tab. The model files travel to you; your image doesn’t travel anywhere.
That’s useful for the things people most often need to copy text from: a letter with an account number, a prescription label, a screenshot with someone’s address. The text box and the image are cleared when you close the tab.
Languages
Six languages are available: English, Spanish, French, German, Italian and Portuguese. Pick the one that matches the text. Running French text through the English model still finds most words but tends to drop or mangle accented letters like é and ç.
What it reads well, and what it doesn’t
It does best on:
- Screenshots, because the text is sharp, straight and high-contrast.
- Printed pages, signs and labels with clear type, photographed straight on in even light and cropped close.
It struggles with:
- Handwriting. Treat any result as a rough draft at best.
- Tables and columns. The text comes out as plain lines, and two columns can get interleaved. Copy one column at a time by cropping.
- Curved or angled pages, like a thick book photographed open.
- Small, blurry or low-contrast text, including text printed over a busy photo.
Tips for cleaner results
Crop before you drop. Cut the image down to the text you want. Less background means fewer stray characters.
Straighten and light it. Hold the phone parallel to the page, in daylight if you can, with no shadow across the lines.
Zoom in on screenshots. Small UI text reads better if you zoom the screen before taking the screenshot.
Check numbers by eye. OCR can confuse 0 and O, 1 and l, or 5 and S. Before you paste an account number, a total or a phone number, compare it with the image.
On your phone
Scanpaper reads text on the phone itself, offline, using the system text recognition (Apple Vision on iPhone, Google ML Kit on Android). After a scan you can copy the text, export it as a TXT file, search your whole library by the words inside your scans, and export PDFs with an invisible text layer that any PDF reader can search. Text recognition is part of Scanpaper Pro, which has a free trial.
Frequently asked questions
Is my image uploaded to extract the text?
No. The recognition engine runs inside your browser tab and reads the image there. The only download is the engine and language model coming to your browser on the first run. Your image and the text from it stay on your device.
Can it read handwriting?
Not reliably. The engine is trained on printed text, so neat block capitals sometimes come through and joined-up handwriting usually doesn't. For handwritten notes, expect to retype most of it.
Which languages does it support?
English, Spanish, French, German, Italian and Portuguese. Pick the language of the text before you run it, since the model for each language knows its accents and letters. Mixed-language text works best with the language that makes up most of it.
Why is the extracted text wrong or jumbled?
The usual causes are a blurry or tilted photo, low contrast, a busy background or a layout with columns and tables. Crop to the text, shoot straight on in good light and try again. The confidence score after each run gives you a rough idea of how clean it was.
Does it work on PDFs?
Not directly; it takes images. Take a screenshot of the PDF page, or export the page as an image, and drop that in. For a scanned PDF you made on your phone, an app with built-in OCR saves the extra step.
Related guides
- OCR Scanner App: Copy Text from Scans, OfflineScanpaper reads the text in your scans on the phone, offline. Copy it, export a TXT, search your library by the words inside, and save searchable PDFs.
- Photo to PDF (free tool)JPG, PNG, WebP or iPhone HEIC photos into one PDF.
- iPhone Scanner to PDF: How to Scan Paper in Notes or FilesUse the iPhone scanner to PDF any paper page: exact steps in Files, Notes and Preview, how edges and filters work, and where the PDF is saved.