How to Copy Text from an Image: OCR Tips for Accurate Results
How text recognition works, why it sometimes gets characters wrong, and simple ways to prepare photos and screenshots so the extracted text needs little fixing.
Published · Updated
What OCR does
OCR, short for optical character recognition, looks at a picture of text and works out which letters are there. It is what lets you copy a reference number from a screenshot, turn a photographed page into editable text, or search a scanned document.
Recognition looks at the shapes of letters, groups them into words and lines, and then uses language patterns to decide between close calls, such as whether a smudged shape is an e or a c. That is why clean, standard printed text is recognised almost perfectly, and unusual text is not.
What it reads well, and what it does not
Works well:
- Printed documents, letters, and book pages
- Screenshots of apps, websites, and messages
- Receipts and bills with clear printing
- Signs and labels photographed straight on
Struggles with:
- Joined-up handwriting
- Decorative or very stylised fonts
- Text on curved surfaces such as bottles
- Small text in a low-resolution image
- Text over busy photos or patterns
Five ways to get more accurate text
- Get closer: the text should be large in the frame. Small text in a big photo loses detail.
- Shoot straight on: angled photos distort letter shapes, particularly at the far edge of the page.
- Light it evenly: shadows across a line of text are a common source of missing or wrong words.
- Crop to the text: remove the desk, the neighbouring page, and images with Crop Image, so the recogniser only sees what matters.
- Remove colour: text on coloured or patterned backgrounds is often read more reliably after converting with Image to Grayscale.
Characters that are easy to confuse
Even with a good picture, some characters look alike: 0 and O, 1, l, and I, 5 and S, rn and m. In ordinary sentences the language patterns usually get these right. In codes, account numbers, serial numbers, and email addresses there is no pattern to help, so always check them against the original before you use them.
Images, PDFs, and scans
For photos and screenshots, use Image to Text. It recognises printed English text and runs in your browser, so the picture is not uploaded.
If you have a PDF, first check whether it already contains real text by trying to select a sentence. If you can, Extract PDF Text gets every word exactly with no recognition errors. If the PDF is a scan, convert its pages to images with PDF to JPG and run each page through Image to Text.
Useful everyday examples
- Copying a booking reference or tracking number from a screenshot
- Turning a photographed recipe, notice, or handout into text you can edit
- Pulling figures from a printed bill into a spreadsheet
- Quoting a paragraph from a slide or a book without retyping it
- Making notes searchable after photographing a whiteboard
Built-in options on your phone
Recent iPhones and Android phones can recognise text inside photos without any extra app. On an iPhone, open the photo and press and hold on the text, or tap the text icon that appears. On Android, Google Lens offers a Text option when you view a photo. These work well for a line or two, such as a phone number.
For longer passages, or when you want to review and edit the whole text before copying, a dedicated tool is easier. Image to Text shows all the recognised text in one box with a Copy button, and works on a computer as well as a phone.
Privacy when the image is sensitive
Photos of ID cards, bank details, and medical letters deserve care. Many online OCR services upload the image to their servers to process it. Docsdom's Image to Text reads the picture inside your browser, so it does not leave your device. Whatever tool you use, avoid running images of other people's documents through services you have not checked.
Try Image to text (OCR)
Free, with no account. It runs in your browser, so your files are not uploaded.
Open Image to text (OCR)