How to copy text from an image without retyping

The image looks clear. The words still refuse to select. If you are wondering how to copy text from image files, the answer is optical character recognition, or OCR: software reads the pixels and turns them into editable characters. You can use a built-in photo feature for a quick line, or a screen capture tool for text inside other apps. This guide takes you from choosing a clean source to checking the pasted result. The examples show what to inspect, so a plausible-looking mistake does not slip into your notes.

Start with the source you actually have

Before installing anything, try selecting the words. A web page may contain real text over an image. A PDF may have a text layer. If ordinary selection works, use it. That avoids a recognition step and preserves characters that a screenshot could blur. OCR is useful when the words are part of the picture itself, rather than just drawn over it.

For a photo or screenshot on a Mac, open the file in Preview and hover over the words. Apple's instructions say, “Drag to select text”. Control-click the selection and choose Copy Text. Apple's Live Text guide also explains the Text Selection tool and language limitations. It is a sensible first option when the file is already open there.

For a dialogue, remote desktop or slide, a region capture avoids making a new image file. The image extraction page explains that route in Lass0. These are different ways into the same task. Choose the one with the fewest conversions between the original and the words you need.

How do I copy text from an image on my screen?

Open the image at a readable size. In Lass0, press the capture shortcut, drag around the text, inspect the result and choose Copy. The default shortcuts are listed in the shortcut documentation, where they can stay in step with the app. On a Mac, screen access must be allowed before the selection can work. Do that during setup, rather than when a meeting is already waiting for you.

Keep the box close to the text, but leave a small margin. Do not cut off the first character on the left or the tails of letters at the bottom. Exclude navigation, unrelated captions and other columns unless they form part of the passage. The recogniser cannot know that the tiny price beside your paragraph belongs to a different product.

Read the preview before moving to another app. If the words are missing, retry with a better source. If the words are right but the order is wrong, select one column at a time. If the grouping is wrong, compare a simpler output format. These are three separate problems, and changing everything at once makes it hard to tell which fix helped.

Choose a paste format before you copy

Plain text suits search fields, tickets and most spreadsheets. Markdown suits notes, issues and documentation where headings and lists matter. Rich text suits an email or word processor that accepts formatted content. A layout rendering is useful when alignment itself carries meaning, but it is not a spreadsheet file or a promise of pixel-perfect reproduction.

Table: format choices for common image-text tasks. These are workflow examples, not performance measurements.

The destination can change the result too. Some editors accept clipboard HTML; others strip it. If a rich-text paste becomes a paragraph, check the target app's paste options before assuming the OCR failed. Keep the recognised words and the destination's formatting behaviour separate in your diagnosis.

A worked example: the notice in a group chat

Imagine a colleague sends a photo of an event notice. You need the venue, date and start time for a calendar entry. This is an illustrative workflow, not a customer story. The tempting move is to capture the whole photo. That may include a sponsor strip, an old date on a poster behind it and a shadow across the address.

Instead, open the original attachment rather than the small chat preview. Crop the selection to the event details, leaving the venue name with its address. Copy as plain text into a local note. Compare the date against the image, then check the time and postcode character by character. Only then transfer them to the calendar. A tidy sentence is not proof that a digit was read correctly.

If the notice says “18:30”, do not quietly turn a questionable “18:8O” into the time you expect. Enlarge the source or ask for a clearer image. The point of OCR is to save typing, not to make an uncertain source look certain. That distinction matters just as much for an order number as it does for an event date.

Why does image text come out wrong?

Small letters, compression, glare and skew can remove the details that distinguish characters. Increasing display size can help when the original contains detail that the current view hides. Enlarging an already blurred thumbnail cannot restore strokes that were never captured. Ask for the original file before spending ages on filters.

The Tesseract image-quality guidance discusses resolution, noise, borders and skew. Its recommendations concern that OCR engine and are useful diagnostic principles, rather than universal settings for every app. For screen work, start with a straight, sharp view and enough space around the words. Avoid aggressive sharpening that creates false edges around letters.

Watch for repeated confusions: a zero and capital O, a one and lowercase l, or a minus sign and a short dash. Names, account references and code deserve closer review than ordinary prose. Spellcheck can catch a misspelt word while completely missing a valid but wrong surname. Use the source, not the spellchecker, as the deciding reference.

Can OCR preserve tables and headings?

Sometimes, but recognising characters and recovering structure are separate jobs. A large bold line may be a heading, a figure label or the first row of a table. Two nearby text blocks may read across the page or down separate columns. Look at the extracted structure as well as the words. A correct amount under the wrong heading is still wrong information.

For a table, include the header and a complete set of rows in the selection. Inspect the pasted cells before sorting or calculating. The image-to-Excel guide covers that review in more detail. For notes with code, headings or lists, use the screenshot-to-Markdown workflow to check syntax as well as content.

There is no need to chase a perfect layout when you only need one address. Choose the smallest useful task, verify it and stop. If you often need text from apps that block selection, download Lass0 and try one ordinary, non-sensitive capture first. You will learn more from checking a real passage in your own destination app than from a generic accuracy claim.

If several people will reuse the extract, keep the checked text with a short source reference rather than asking each person to recognise the same image again. That reduces repeated correction work and gives later questions a clear route back to the original. Keep any uncertain characters visibly unresolved until someone can verify them.

Questions, answered

Can I copy text from a JPG or PNG?

Yes, if an OCR tool can read the displayed text. Open the original at a useful size and check the recognised characters before sharing them.

Does making an image larger improve OCR?

It can expose detail hidden by the current view, but enlarging a blurred thumbnail does not recover lost information. Start with the original file.

Can image-to-text tools read handwriting?

Support varies by engine, language and handwriting. Test a representative sample and review every important field; printed-text support does not establish handwriting accuracy.