Image to Text: What OCR Can Really Read from a Screenshot or Photo

A screenshot of a receipt, a photo of a business card, a scanned page, a slide with a paragraph you need to quote. Image to text looks like a simple copy button, but OCR is a prediction. It works best when the text is printed, straight, large enough and evenly lit. It struggles when the text is small, angled, handwritten or buried in a busy background.

The tool outputs a text file and shows the text in a box. It does not preserve the layout, the images or the colours. For a searchable PDF that keeps the page, use OCR PDF instead.

What kinds of images work well

Screenshots are the easiest. The text is already digital, sharp and high-contrast. A screenshot of a chat, a web page or a slide is an ideal OCR input.

Scanned documents at 300 DPI are also good, provided the page is straight and the text is not too small. A clean printed page with a white background is the classic success case.

Photos of documents can work if the light is even and the camera is held straight. A photo taken at an angle, with a shadow from the hand, is much harder. A scanner app that crops and straightens the page gives better results than a raw camera photo.

Business cards, receipts and labels are possible but variable. A receipt on glossy paper can have reflections; a business card with a logo behind the text can confuse the model.

What defeats OCR

Handwriting is the big one. A clear signature is not the same as a clear sentence. Cursive, mixed Arabic and Latin script, and unusual letter forms produce errors. If the handwriting matters, transcribe it manually or use a specialised handwriting model.

Small text is another. A paragraph photographed from a distance may be readable to a human who zooms in, but the OCR engine may not have enough pixels per character. Crop the region and use a larger image.

Low contrast, shadows and glare reduce accuracy. So do decorative fonts, text over images, and coloured text on a coloured background. A menu photographed under warm restaurant lighting is a hard case.

Tables and multi-column layouts are difficult because OCR reads in a visual order. It may read across columns instead of down each column, producing sentences that jump between topics.

Choose the language

Select Arabic for Arabic text, Indonesian for Indonesian, English for English. The language setting changes the character set and the word model. Choosing the wrong one produces strange output: Latin letters recognized as Arabic, or Arabic words broken into punctuation.

If an image mixes languages, OCR will favour the selected one. A business card with an English name and an Arabic address may need two passes. Check the parts in the other language manually.

The language setting is not the interface language. If your browser is in French but the image is in English, choose English.

Prepare the image

Crop to the text you need. A screenshot with a large empty area around a short paragraph gives the engine more background to ignore. Cropping also reduces processing time.

Straighten the text. A small rotation can be corrected with an image editor or a scanner app. OCR can handle a slight tilt, but not a page photographed at 30 degrees.

Increase contrast if the text is faint. Do not overdo it; thin strokes can disappear. For a photo with a shadow, use the editor’s brightness and contrast, or reshoot with better light.

If the image is a screenshot, use the original file, not a photo of the screen. A photo of a monitor adds moiré, reflections and perspective distortion.

Run the OCR

Open the Image to Text tool. Select the image. Choose the language. Run the recognition.

The text appears in a box and can be downloaded as a .txt file. Read the result before using it. OCR does not know which words are names or numbers; it outputs what it thinks it sees.

For a long document, use the PDF OCR tool. For a single screenshot or receipt, the image tool is faster.

Proofread the result

Check names, numbers, dates, reference codes and currency amounts first. These are the fields where a misread character causes real damage. 0 and O, 1 and l, 5 and S are common confusions.

Check the beginning and the end of the text. OCR sometimes drops the first or last line, especially if it is close to the image edge. Check that the text is in the right order; columns can scramble it.

If the text is for a quote, compare the wording with the image. OCR may change punctuation, capitalisation or a small word. A quote with a missing “not” can change the meaning.

Formatting and line breaks

The text output preserves line breaks as the engine sees them. A paragraph may be split into many short lines. When you paste it into a document, you may need to join the lines manually.

Do not assume the output has paragraphs. OCR does not understand the document structure. It recognizes text, not headings, lists or tables.

If you need a table, copy the numbers into a spreadsheet and check each row. OCR may preserve the numbers but lose the column alignment.

When to use OCR PDF instead

Use OCR PDF when you need the page to remain a PDF and the text to be searchable behind the image. Use Image to Text when you only need the text, for example to paste it into an email, a note or a translation tool.

The image tool does not create a searchable PDF. If the recipient needs to search the original page, use the PDF tool.

Screenshots versus photos

A screenshot is a direct capture of pixels that were already digital. OCR often gets nearly every character right. A photo of the same screen adds noise, reflections and perspective; the accuracy drops.

If you can take a screenshot, take it. On a phone, use the screenshot buttons; on a computer, use the system screenshot tool. Crop the screenshot to the text area and run OCR.

A photo of a paper document can be excellent if it is flat, evenly lit and sharp. Use the rear camera, avoid the flash, and shoot from directly above. A scanner app is even better because it crops and straightens the page.

Receipts and business cards

Receipts often use small, condensed fonts and glossy paper. A reflection can erase a line. Photograph the receipt on a flat surface with light from the side, not directly above, to reduce glare. If the receipt is long, take several close-up photos rather than one distant photo.

Business cards can have decorative fonts and text over a logo. OCR may read the name and phone number but mangle the address. Check the phone number digit by digit; a single wrong digit is worse than a missing line.

Tables and columns

OCR reads text in a visual order, but it does not understand a table. A two-column page may be read across both columns, mixing sentences. A table may lose its rows or columns even if every number is recognized.

If the table is the important part, do not rely on OCR alone. Copy the values into a spreadsheet and verify each cell against the image. For a menu or price list, check the prices separately.

How to test accuracy

Pick a line with a name, a line with a number and a line with punctuation. Compare the OCR text with the image word by word. If those lines are correct, the result is probably usable for searching.

If the result is poor, change the input, not the language setting. A sharper crop, better light or a higher-resolution image usually helps more than running OCR again on the same pixels.

Common mistakes

• Photographing a screen instead of taking a screenshot.
• Using a photo with a shadow across the text.
• Choosing the wrong language.
• Expecting OCR to preserve paragraphs and columns.
• Trusting a phone number or amount without checking it.
• Using OCR for handwriting and assuming it is correct.
• Deleting the original image before checking the text.

If the text is still unreadable, try a better source image rather than a different tool; a blurry image is the limit. Keep the original until the extracted text has been checked.

Privacy

The image is processed in the browser. OpenToola does not upload it and does not keep a copy. The text is generated locally and downloaded to your device.

That is useful for receipts, identity documents and client screenshots. It does not make a shared computer safe. Delete the image and the text file after use if they contain sensitive information, and avoid OCR on a public machine.

Tools mentioned

• Image to Text
• OCR PDF
• Compress Image
• OpenToola Guides