Screenshot to editable text
Paste a screenshot, get the text back. The recogniser runs in your own browser, so the image is never uploaded to an OCR service, and the result lands in an editable box you can fix, copy or download.
What does this screenshot-to-text tool do?
It reads the text out of a screenshot or any other image and hands it back as editable text you can copy, fix or download as a file. The recognition runs in your own browser in a Web Worker, so the image is never uploaded and nothing is sent to an OCR service. Add an image or paste one from the clipboard, then edit the result in place.
Your screenshot never leaves this device. The OCR engine is served from this site and runs in a Web Worker on your own machine. No image is sent to an OCR service, no request carries one, and closing the tab discards everything.
Empty state. No screenshots yetAdd one above, paste one from the clipboard, and the text will appear here.
English is the model that ships with this page. It is stored by your browser after the first run, so later recognitions start faster.
Automatic suits a full screenshot. Pick one line for a single strip of text, or sparse text when the words are scattered across the image.
Getting a better result
- Capture at full size rather than scaling the screenshot down.
- Crop to the text you need โ a small sharp region beats a full desktop.
- High contrast helps most: dark text on a light background reads best.
- Handwriting and heavily filtered images are still read poorly by every OCR.
How to turn a screenshot into editable text
Add the screenshot
Drop a PNG, JPG or WebP onto the area above, use the file picker, or paste the screenshot straight from the clipboard. Several images can be added at once, up to 25 MB each.
Wait for the recognition
The OCR engine runs in a Web Worker on your own device. A progress bar shows the language model loading on the first run and then the text being read.
Edit and take the text
The recognised text lands in an editable box. Correct any mistakes the recogniser made, then copy a single result or everything at once, or download the text as a file.
Where the recognition actually happens
OCR is usually sold as an API: you POST a picture and get text back, which means somebody else has a copy of your picture. This page does it the other way round. The recogniser, its language model and its core engine are all served from this site and run inside a Web Worker on your own machine, so the image never becomes part of a request.
That is also why the first run takes a moment: the language model is downloaded and stored by your browser once, then reused. The progress bar under the file area says what stage the worker is at, and everything already on the page keeps working while it loads.
Why the result is editable
OCR is good, never perfect. Small type, tight leading, icons next to words and low contrast all produce mistakes, which is why the text lands in an editable box instead of a finished document. Correct what is wrong, then copy it or download it as a plain text file.
If the text you need is inside a PDF rather than an image, the PDF accessibility checker reads a document for its structure, the text fixer cleans up what you get out, and the text tools page counts it and converts it to the case you want.
Common questions
Is my screenshot uploaded?
No. The image is decoded and recognised by your own browser, using an OCR engine that is served from this site and runs in a Web Worker on your device. No part of the picture is sent anywhere, and closing the tab discards it.
How accurate is the text recognition?
Good on clean screenshots โ UI text, documents, code and anything with real contrast โ and noticeably worse on handwriting, heavy filters, low-contrast photos and very small type. The output is an editable box rather than a finished document precisely so the mistakes can be corrected by hand before you copy it.
What languages does it read?
English, which is the language model that ships with this page. The model is stored by your browser the first time you run a recognition, so later runs start faster and still work without sending your image anywhere.
What image formats can I use?
PNG, JPG and WebP โ the formats a screenshot is normally saved as. Take the screenshot at full size rather than a thumbnail, and crop to the text you want: recognition works better on a small, sharp region than on a full desktop with text in it.
Can it extract text from a PDF?
Not on this page โ this tool takes images. If the PDF has real text in it, the PDF accessibility checker at /pdf-accessibility-checker reads the same file for its structure, and if the PDF is a scan you need an OCR tool built for multi-page documents.