Get the text out of an image
Read the words in a photo, a screenshot or a scan and hand them back as text you can edit.
Runs in your browser · nothing uploaded
Add your files. Nothing is uploaded — the work happens in this tab.
Drop JPG, PNG, WebP, GIF, BMP or TIFF images here
or
JPG, PNG, WebP, GIF, BMP or TIFF images · up to 50 MB each
You can also paste a file from your clipboard
Reading the text
Every image you add, with no limit on how many
Your processor is doing this, so there is no per-page cost for us to limit. Most free OCR stops at 50 pages because every page costs the site server time.
Languages in the image
- No published language matches “”. Ten are published here; the rest are not built into this site yet.
Pick as many as the page needs — there is no three-language limit here. Each extra language makes reading slower.
Language sizes load with the page. Nothing is downloaded until you start.
What you get
The words on their own, with a line between each image.
A word the reader is not sure of is left out rather than guessed at, and stays in the picture. A image that is mostly unclear keeps no text at all and is named in the summary. A search that finds the wrong document is worse than one that finds nothing.
Add a file first. These settings are used when you start.
Runs on your device · nothing uploaded
How to get the text out of an image
- Drop your images inOne or several — a photo of a page, a screenshot, a receipt.
- Pick the languagesChoose the languages in the image. The first use of a language downloads its data once, and the size is shown before it does.
- Read the text and take itSelect Read the text. Each image is read in turn on your own processor; copy the result straight off the page, or download it as a .txt file.
Questions
Which image formats work?
JPG, PNG, WebP, GIF and BMP everywhere. TIFF only where your browser has a decoder for it, and Chrome and Firefox do not — a picture the browser cannot decode is named in the summary rather than skipped quietly.
Will it read handwriting?
Rarely, and not reliably. This is built for printed text. Handwriting needs a different kind of model, and pretending otherwise would only produce confident nonsense.
How do I get the best result?
A straight, evenly lit, in-focus image at a reasonable size. A photo taken at an angle, or a screenshot scaled down to thumbnail size, loses the detail the reader needs.
Is my screenshot uploaded anywhere?
No. The image is decoded and read inside this tab. Screenshots often contain more private information than people expect, which is exactly why this tool does not send them anywhere.