Does OCR recognize Lithuanian letters?

Yes, if Lithuanian is selected. With another language set, ą, č, ę, ė, į, š, ų, ū, ž are often replaced with wrong characters.

Can it read handwriting?

Basic OCR handles it poorly. Neat, separated block letters work better than cursive, but expect many errors.

Is my image sent to a server?

No. Recognition happens in your browser, so the image stays on your device.

Can I trust the result without checking?

Not always. Always verify numbers, dates, and names, since one misread digit can change the meaning of an amount.

How to Extract Text From an Image or Screenshot (OCR)

Text is often stuck inside an image: a photographed receipt, a scanned document, a screenshot, or a slide. Optical character recognition (OCR) can turn it into editable text. This guide explains how it works and what to do to get an accurate result.

What OCR is

OCR (optical character recognition) is technology that looks for letter shapes in an image and converts them into real text. To a computer, a photo is just a grid of pixels, so without OCR you cannot select, copy, or search the words in it.

The result is not magic: the software guesses each character based on its shape and a language dictionary. That is why a clear, well-lit photo gives a far better result than a tilted, blurry one.

How to extract text step by step

Open the image to text tool and choose your picture (JPG, PNG, WebP or similar).
Select the document language. Lithuanian text with characters like ą, č, ę, ė, į, š, ų, ū, ž is recognized more accurately when the language is set correctly.
Start recognition and wait: the first time, language data may need to be downloaded, so it can take longer.
Review the result and fix mistakes, especially numbers, dates, names, and amounts.
Copy the text or save it to a file.

Preparing the image for better results

Source quality matters more than anything else. A few simple steps often improve the result more than any setting.

Shoot straight from above so lines are not slanted or distorted.
Use even lighting without shadows or flash glare.
Characters should be large enough: if the image is very small, enlarge it before recognition, and if it is huge, scale it down so processing is faster.
Crop unneeded edges so the text stands out clearly from the background.
Save screenshots as PNG rather than heavily compressed JPG: artifacts around letters confuse recognition.

What works well and what does not

Source	Expected accuracy
Screenshot with a printed typeface	Usually very good
Well-scanned printed document	Good, still needs review
Receipt photographed with a phone	Moderate, depends on light and paper condition
Handwriting	Often poor, especially cursive
Heavily stylized or decorative fonts	Unreliable

Where OCR is most useful

Typical examples: receipt and invoice data you need to move into a spreadsheet, passages from scanned books or articles, text from slides or screenshots that you want to quote, and photos of old documents that you want to make searchable.

If you need several pages, it is best to process them one at a time and combine the results by hand. Watch out for tables: OCR often reads the words correctly but breaks the column structure, so tabular data has to be tidied up manually.

Privacy

Photos often contain sensitive data: personal ID numbers, account numbers, medical notes. Many online OCR services upload your file to their server. The Konvertavimas.lt tool runs recognition in your browser, so the image is never sent anywhere.

When you are done, clear unneeded text from your clipboard and downloads folder.

What to do if the result is poor

First, check the selected language. A common mistake is recognizing Lithuanian text with English selected, which turns accented letters into plain ones or odd symbols.

If the text is still garbled, try increasing the contrast or rotating the image the right way up. If that does not help, retake the photo under better conditions. For long results, a word counter helps you sanity-check the output, and recurring wrong characters can be fixed with find and replace.