English
How to convert an image to text
An image to text converter reads the letters in a picture and gives them back as text you can edit, search and paste. The technical name is OCR, optical character recognition. Here is the short version.
- Choose an image, drop it on the box, or paste a screenshot with Ctrl+V.
- Check the language chips. English and Indonesian are on by default; switch on any other language that appears in the picture.
- Press Extract text. A progress bar shows the image being read.
- Fix any slips in the box, then copy the text or download it as a .txt file.
The first run is slower because the browser has to fetch the text engine and the language data. After that it is cached, and a screenshot usually takes 5 to 20 seconds depending on your phone or laptop.
Why the text sometimes comes out wrong
OCR does not understand words, it matches shapes. So anything that blurs a shape causes errors: a shaky photo, a page taken at an angle, a shadow across the paper, a tiny font. Most bad results come from the picture, not the tool.
| Problem in the picture | What you see in the text | Fix |
|---|---|---|
| Blurry photo | "rn" for "m", "l" for "I", missing letters | Retake it with the page filling the frame |
| Tilted page | Lines merged or cut in half | Photograph straight on, or rotate the image first |
| Wrong language | Accents missing, words broken into pieces | Select the right chip, or add a second one |
| Small image (under about 800 px tall) | Random symbols | Keep the cleanup option on so it is enlarged first |
| Coloured or dark background | Gaps, lost lines | Cleanup option, or crop to the text area |
The Confidence number is the engine's own estimate. Above 85% you will usually only fix a word or two. Below 70% it is often quicker to retake the photo than to correct the text.
Which layout setting should I pick?
Leave it on Automatic for normal pages. Switch to One block of text for a single paragraph or a cropped section, since that stops the engine from guessing at columns that are not there. Use A single line for a number plate, a serial number or a caption. Use Scattered text for photos of signs, shelves or packaging, where the words sit in different places.
Can it read several languages in one image?
Yes, but select only the ones you need. Each extra language makes the first load heavier and slows the reading slightly, and it can introduce wrong guesses. A menu in Indonesian with English dish names is a good case for two chips. A page in one language needs one.
Copying text from a screenshot into Word
Take the screenshot, press Ctrl+V on this page, extract, then copy. Tables and multi-column layouts come out as plain lines in reading order, so expect to rebuild the table in Word. If the lines break in the middle of sentences, tick Join broken lines into paragraphs and the text flows again. Words split by a hyphen at the end of a line are rejoined too.
What about handwriting and PDFs?
Handwriting is outside what this engine does well. Clear block capitals sometimes work; joined-up writing almost never does. For a scanned PDF, export each page as an image first (the PDF to JPG tool does this) and drop those images here. A PDF you can select text in already has its text, so you do not need OCR at all.
Is it private?
The image is read inside your browser. What is downloaded is the engine and the language files, not the other way round. That makes it a reasonable choice for receipts and notes. If a document is truly sensitive, you can still go offline once the page has loaded and the language data is cached.
Frequently asked questions
How do I extract text from an image?
Add the image, check that the language matches, and press Extract text. A one-page screenshot usually takes 5 to 20 seconds, then you can copy the text or download it as a .txt file.
Is image to text free, and are my pictures uploaded?
It is free and your picture stays on your device. The page downloads the text-reading engine and the language data you pick, but the image itself is never sent anywhere.
Can it read Indonesian as well as English?
Yes. English and Indonesian are selected by default, and you can add Malay, Javanese, Sundanese and 18 more. Select every language that appears in the image.
Why is the text wrong or full of strange characters?
Usually the photo is blurry, tilted, low resolution or the wrong language is selected. Retake the photo straight on in good light, keep the cleanup option on, and make sure the right language chip is highlighted.
Can I copy text from a screenshot into Word?
Yes. Paste the screenshot with Ctrl+V, extract the text, then copy it into Word or Google Docs. Tables and columns come out as plain lines, so expect to reformat them.
Does it read handwriting?
Not reliably. The engine is built for printed text such as books, receipts, signs and screenshots. Neat block capitals sometimes work, but cursive handwriting usually comes out wrong.
How accurate is OCR on a phone photo?
On a sharp, well-lit printed page it is often right for most words, and the Confidence figure tells you how sure the engine was. Below about 70% expect to fix many words by hand.
Can I convert several images at once?
Yes. Select or drop several images and they are read one after another into one text, each under its file name. Download the .txt when it finishes.