How to Get Text Out of a Photo or Screenshot (Free OCR Guide)

Your coaching centre shares the test schedule as a photo in the WhatsApp group. Or a client sends a screenshot of their address, and now you need to type it into an invoice. You could retype everything by hand, squinting at the screen and hoping you don't swap two digits in a pincode. Or you could let your computer read the text for you.

That second option is called OCR, and it's free and quick when the image is decent. This guide shows you how to use it, how to tell in advance whether your photo is good enough, and where it still lets you down.

What OCR does (and what it doesn't)

OCR stands for optical character recognition. A photo or screenshot is just a grid of coloured pixels. Your eyes see words in it, but your computer doesn't, which is why you can't select or search the text. An OCR engine looks at the shapes in the image, works out which letters they most likely are, and gives you back plain, editable text.

It works best on clean, printed text: screenshots, typed notices, book pages, receipts, printed forms. It struggles with handwriting, fancy decorative fonts, blurry photos and pages shot at a steep angle. Our Image to Text tool uses Tesseract, a well-known open-source OCR engine trained on printed text, so be realistic: if your image is a handwritten letter, expect to retype most of it.

One thing OCR does not give you is the layout. You get the words, in reading order, as plain text. Tables, columns and bold headings won't come across as formatting.

How to extract text from an image with Image to Text

1. Open Image to Text and drop your image onto the box that says "Drop an image here or click to browse". JPG, PNG and WebP all work. You'll see a preview with the file name.
2. Under Language of the text, pick the language your image is written in. English is the default. The other choices are Hindi, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Chinese (Simplified), Japanese and Korean. You can choose one language per scan.
3. Click Extract Text. The first time, you'll see messages like "Loading OCR engine..." and "Downloading language model..." because your browser fetches the engine and language data once and caches it. Later scans are faster.
4. When it finishes, the text appears in an editable box along with a confidence score (0 to 100) and a word count. Fix any mistakes right there in the box.
5. Click Copy to put the text on your clipboard, or Download .txt to save it as a text file named after your image.

The recognition runs inside your browser, so the image itself isn't uploaded to our server. That makes it a reasonable choice for things like a screenshot of a bank address or an office notice. If your document is a scanned PDF rather than an image, use PDF OCR instead, which does the same job page by page.

Worked example: is your photo sharp enough to read?

The single biggest factor in OCR accuracy is how many pixels each letter gets. Tesseract's own documentation says it works best on images of at least 300 DPI (dots per inch). Here's what that means in practice.

Text size is measured in points, and one inch has 72 points. So the height of a 12-point font in pixels is 12 ÷ 72 × DPI:

• At 300 DPI: 12 ÷ 72 × 300 = 50 pixels
• At 150 DPI: 12 ÷ 72 × 150 = 25 pixels
• At 100 DPI: 12 ÷ 72 × 100 = about 16.7 pixels

Now apply it to a phone photo. Say your camera takes 12-megapixel photos at 4032 × 3024 pixels, and you photograph an A4 page (297 mm, or about 11.69 inches, tall) so the page fills the long side of the frame. That gives you 4032 ÷ 11.69 ≈ 345 pixels per inch, comfortably above 300. Normal 10-point print would be about 10 ÷ 72 × 345 ≈ 48 pixels.

Stand back so the page only fills half the frame, and you drop to about 172 pixels per inch. That same 10-point text shrinks to about 24 pixels. It may still read, but small print, footnotes and numbers start to suffer. The lesson is simple: fill the frame with the text you need. Getting closer helps more than any setting.

Screenshots are usually easier. The text is perfectly sharp and straight, with no shadows. A screenshot taken at normal zoom tends to read well, while one cropped from a zoomed-out page can have letters only a few pixels tall.

Tips for cleaner results

• Shoot straight on. Hold the phone parallel to the page. Tilted lines and perspective distortion confuse the engine. If a photo is rotated, straighten it first so the text lines are horizontal.
• Use even light. A shadow from your hand or phone across half the page can wipe out whole lines. Daylight near a window usually works better than an overhead bulb.
• Crop to the text. Borders, logos and backgrounds can turn into random characters. Leave a small margin around the text instead of cutting right up to the letters.
• Pick the right language. A Hindi notice read with English selected will come out as gibberish. If a page mixes Hindi and English, consider scanning it twice with each language and taking the good parts from each.
• Watch the confidence score. The tool's guide suggests proofreading carefully if the score is below about 80. Even a high score isn't a guarantee, though.

Mistakes to avoid

• Trusting numbers without checking. OCR commonly confuses 0 and O, 1 and l, 5 and S, or 8 and B. For phone numbers, account details, amounts, GSTINs and roll numbers, always compare with the original image character by character.
• Expecting handwriting to work. Neat block capitals sometimes come through. Cursive usually doesn't.
• Thinking the tool has frozen. The first scan downloads the engine and language model, so it takes longer. Give it a moment before reloading.
• Pasting straight into a final document. OCR text often has odd line breaks where the original lines ended. If the source was in ALL CAPS, our Case Converter can turn it into Sentence case so it's easier to read and edit.

For text that's already inside a normal (not scanned) PDF, you don't need OCR at all. Our post on extracting text from a PDF explains how to tell the difference.

Frequently Asked Questions

Is my image uploaded anywhere?

No. Image to Text runs the OCR engine inside your browser, so the image stays on your device. The engine and language files are downloaded from a public CDN the first time you use it, which is why you need an internet connection for the first scan.

Can it read Hindi text?

Yes, Hindi is one of the 12 languages you can choose under "Language of the text". It works on printed Hindi. Handwritten Hindi will be unreliable, just like handwritten English.

Why does my result have strange symbols?

Usually because of blur, shadows, a busy background, the wrong language setting, or a border that got read as characters. Crop tighter, retake the photo straight on in better light, and check the language.

Can I get the text out of a photo on my phone without a website?

Often, yes. iPhones have a built-in feature Apple calls Live Text for copying text from photos, and many Android phones offer something similar through Google Lens. A browser tool is handy when you're on a laptop, need a .txt file, or want to pick a specific language.

Does the tool keep the table or column layout?

No. You get plain text in reading order. For tables, you'll usually need to paste the text into a spreadsheet and tidy up the columns yourself.