How to extract text from an image

Extract text step by step

1. Open Image to Text.
2. Add a JPG, PNG or WebP image that contains text.
3. The first time, the recognition engine downloads to your browser.
4. When the text appears, correct anything it misread and copy it.

How to get accurate results

• Take the photo straight on, not at an angle.
• Make sure it is sharp and evenly lit, with no glare across the words.
• Crop to just the text with Crop Image. Less clutter means fewer mistakes.
• Low contrast text is harder to read. Converting to black and white sometimes helps.
• Screenshots usually give the best results because the text is perfectly sharp.

What OCR cannot do well

This version works best with English printed text. Handwriting is unreliable, and tables lose their column layout, so you may need to tidy the result up.

Text from a scanned PDF

Turn each page into an image first with PDF to JPG, then run the pages you need through Image to Text.

What works well

• Screenshots of web pages, documents and chats, where the text is sharp and evenly lit.
• Photos of printed pages, such as a book page or a letter, taken straight on in good light.
• Receipts and labels, if they are flat and well lit.
• Signs and posters in clear, common fonts.

Prepare the image for the best result

1. Crop to only the text with Crop Image, so background clutter does not confuse the reader.
2. Make sure letters are large enough. Text that is roughly 20 pixels tall or more reads far better than tiny text.
3. Use even light and avoid shadows and glare across the words.
4. Hold the camera straight on, parallel to the page. Angled photos bend the lines and cause mistakes.
5. If the text is pale on a similar background, try Black and White to raise the contrast.

Typical mistakes to look for

Even a good result needs a quick proofread. Character recognition most often confuses similar shapes: the letter O and the number 0, a lowercase l and the number 1, the letters r and n together with m, and the letter S with the number 5. Check any numbers, prices, names and email addresses by eye against the image.

Line breaks can also differ from the original, because the text is read line by line. Join broken lines or remove extra spaces after you paste it.

Working with many pages

For a multi-page scan, convert the PDF with PDF to JPG, then read one page at a time. Paste each result into the same document in order. For long documents, it is worth checking each page right after reading it, so you catch problems early.

Limits to know

This version is built for printed English text. Handwriting, decorative fonts, curved text and photos taken at steep angles are unreliable. Tables lose their columns, so rebuild them by hand if you need them.

Frequently asked questions