How do I extract text from an image or screenshot for free?

Pick your document language, then drop a JPG, PNG, WebP, or BMP file into the tool above. Tesseract runs in your browser, recognises the characters, and puts the result in an editable box you can correct, copy to the clipboard, or download as a plain .txt file.

Is my image uploaded to a server during OCR processing?

No. Recognition runs locally through Tesseract compiled to WebAssembly, so the image is read from your device into browser memory and never transmitted. That matters when the file is an identity document, a payslip, a medical form, or a signed contract.

Which languages can the OCR engine read?

Eight are available: English, Hindi, Spanish, French, German, Portuguese, Italian, and Arabic. Choose the language before uploading, because the engine loads a model trained for that script. Picking the wrong one is the most common cause of garbled output.

How accurate is the text recognition?

Clean screenshots and sharp 300 DPI scans of printed text usually come back with very few errors. Accuracy drops with low resolution, camera blur, skewed angles, heavy shadows, coloured or textured backgrounds, and unusual display fonts. The tool reports a confidence score so you can judge each result.

Can it read handwriting or cursive notes?

Not reliably. Tesseract is trained on printed and digital typefaces, so neat block capitals sometimes work while joined-up cursive generally does not. For handwritten notes, expect to treat the output as a rough first draft and correct it in the editable box.

How do I get a table into Excel rather than plain text?

Plain OCR flattens a table into lines of text and loses the column structure. Use our Image to Excel converter instead, which reconstructs rows and columns and exports a spreadsheet file. Clean, ruled tables convert well; merged cells and complex layouts do not.

Can I extract text from a scanned PDF?

Yes, two ways. Convert the pages to images with our PDF to JPG tool and drop them here one at a time, or use the OCR PDF tool, which processes every page and writes an invisible text layer back into the PDF so the document itself becomes searchable.

Why does the first run take longer than the ones after it?

The language model has to download once before recognition can start, then your browser caches it. Later images in the same language begin almost immediately. Switching language triggers a fresh model download for that script.

Image to Text (OCR)

Extract text from photos, screenshots and scans in 8 languages — OCR runs in your browser, nothing is uploaded.

100% Private — Runs entirely in browserInstant calculation & resultFree, no sign-up

Document Language:

Choose an image to extract text or drag & drop100% private — extracts text from photos & screenshots inside browser

or import from

OCR, or optical character recognition, turns a picture of text into text you can select and edit. Drop in a screenshot, photo, or scan, choose the language, and this tool reads the characters entirely inside your browser. Nothing is uploaded, and the result arrives in an editable box ready to copy or download.

Why pull text out of an image at all?

Text gets trapped in pictures constantly. An error dialog you need to search for. A page of a library book you cannot borrow. A whiteboard after a meeting. A posted invoice, a receipt, a form filled in years ago and filed as a scan. Retyping any of it is slow and introduces its own mistakes, particularly with long reference numbers. Common jobs people bring here include copying error messages and code snippets out of screenshots, digitising book passages and study notes, transcribingwhiteboard photos into minutes, and lifting text from aforeign-language document so it can be pasted into a translator.

How in-browser OCR works

The tool runs Tesseract compiled to WebAssembly. When you choose a language, a trained model for that script downloads once and is cached by your browser. Recognition then happens on your own processor: the engine finds text lines, works out baselines and character boundaries, and matches shapes against the model, reporting progress as it goes. Because the whole pipeline is local, there is no upload wait, no queue, and no per-file quota — and it keeps working on a flaky connection once the model is cached.

ToolJiffy compared with cloud OCR and desktop scanners

Where local recognition wins, and where it genuinely does not.
	ToolJiffy	Cloud OCR service	Desktop scanner software
Image leaves your device	Never	Yes, uploaded	Never
Account or sign-up	None	Usually required	Licence
Page or credit limit	None	Free credits, then paid	None
Install needed	No	No	Yes
Languages here	8	Often 100+	Varies by product
Handwriting	No	Sometimes	Sometimes
Price	Free, unlimited	Free tier, then paid	Paid

Getting the best possible accuracy

Set the language first. An English model reading Hindi or Arabic produces nonsense, not a near miss.
Feed it resolution. Around 300 DPI for scans, and a full-size screenshot rather than a shrunken one.
Straighten the shot. Photograph pages square-on in even light, avoiding shadow across the text.
Crop to the text. Removing page edges, staples, and background clutter cuts stray characters.
Check the confidence score, then proofread numbers, punctuation, and anything that must be exact.

What OCR handles well, and what it does not

Being straight about the ceiling saves you time. Reliable: printed books and articles, typed letters, invoices and receipts printed on a laser or thermal printer, application forms, and screenshots of ordinary interface text. Workable with care: phone photos of paper, faxed or photocopied pages, small print, and neat block-capital handwriting. Poor: cursive handwriting, decorative or script display fonts, text laid over busy photographs, watermarked pages, very low-resolution images, and anything strongly skewed or curved, such as a photograph of a thick book's inner margin. Multi-column layouts read as text but the reading order can jump between columns, so check the sequence before you rely on it.

Privacy when the image is a document

Passports, driving licences, bank statements, and medical forms are exactly the files people casually drop into free cloud OCR sites. Doing that hands a third party a full copy of the document and its extracted text, and leaves you depending on their retention and access policy. Local recognition avoids the question: there is no upload, no server-side copy, and nothing to request deletion of afterwards.

What to do with the extracted text

The output box is editable, so fix any misread characters before you copy. From there,count the words if you are working to a limit,fix the capitalisation if a scan came back shouting in block capitals, or send a structured table to Image to Excel instead. For multi-page documents, theOCR PDF tool is the better route because it makes the original PDF searchable rather than producing loose text files.

Frequently Asked Questions

How do I extract text from an image or screenshot for free?
Pick your document language, then drop a JPG, PNG, WebP, or BMP file into the tool above. Tesseract runs in your browser, recognises the characters, and puts the result in an editable box you can correct, copy to the clipboard, or download as a plain .txt file.
Is my image uploaded to a server during OCR processing?
No. Recognition runs locally through Tesseract compiled to WebAssembly, so the image is read from your device into browser memory and never transmitted. That matters when the file is an identity document, a payslip, a medical form, or a signed contract.
Which languages can the OCR engine read?
Eight are available: English, Hindi, Spanish, French, German, Portuguese, Italian, and Arabic. Choose the language before uploading, because the engine loads a model trained for that script. Picking the wrong one is the most common cause of garbled output.
How accurate is the text recognition?
Clean screenshots and sharp 300 DPI scans of printed text usually come back with very few errors. Accuracy drops with low resolution, camera blur, skewed angles, heavy shadows, coloured or textured backgrounds, and unusual display fonts. The tool reports a confidence score so you can judge each result.
Can it read handwriting or cursive notes?
Not reliably. Tesseract is trained on printed and digital typefaces, so neat block capitals sometimes work while joined-up cursive generally does not. For handwritten notes, expect to treat the output as a rough first draft and correct it in the editable box.
How do I get a table into Excel rather than plain text?
Plain OCR flattens a table into lines of text and loses the column structure. Use our Image to Excel converter instead, which reconstructs rows and columns and exports a spreadsheet file. Clean, ruled tables convert well; merged cells and complex layouts do not.
Can I extract text from a scanned PDF?
Yes, two ways. Convert the pages to images with our PDF to JPG tool and drop them here one at a time, or use the OCR PDF tool, which processes every page and writes an invisible text layer back into the PDF so the document itself becomes searchable.
Why does the first run take longer than the ones after it?
The language model has to download once before recognition can start, then your browser caches it. Later images in the same language begin almost immediately. Switching language triggers a fresh model download for that script.