What is Image to Text (OCR)?
Optical character recognition turns a picture of text into text you can select, search, and edit. Photograph a page, screenshot a document you cannot copy from, or scan a receipt, and OCR reads the shapes back into characters. This tool runs Tesseract — the most widely used open-source OCR engine, originally developed at HP and maintained by Google for over a decade — compiled to WebAssembly so it runs inside your browser.
That last detail matters more than it sounds. Almost every free OCR site uploads your image to their servers, where it may be stored, logged, or used for training. Documents people run through OCR are exactly the documents you would not want uploaded: contracts, invoices, ID cards, medical letters, bank statements. Here the recognition happens on your own device — only the language model itself is downloaded, once, and then cached.
Fourteen language options are available including English, Hindi, Spanish, French, German, Arabic, Chinese, Japanese, and Korean, plus a combined English + Hindi mode for mixed documents. The result comes back with a confidence score, is fully editable before you copy it, and can be downloaded as a plain text file.
Use Cases
Here are the most common ways people use Image to Text (OCR) every day.
Copying Text From a Screenshot
Text in an image, a PDF that will not select, a video frame, or an application that blocks copying — all of it can be recovered with OCR. This is the single most common reason people reach for an OCR tool, and screenshots are the ideal input: the text is crisp, evenly lit, and already at screen resolution, which is exactly what the recogniser wants. Paste directly from your clipboard rather than saving a file first.
Digitising Printed Documents and Notes
Photograph a page from a book, a printed handout, or your own handwriting-free notes, and get editable text back. Printed text at reasonable resolution recognises very well. Handwriting does not — Tesseract is trained on printed type, and cursive in particular will come back as nonsense. For handwritten notes you need a purpose-built handwriting model.
Extracting Data From Receipts and Invoices
Pull the line items and totals off a receipt into a spreadsheet without retyping. Accuracy depends heavily on the photograph: hold the camera parallel to the paper, get even light with no shadow across the text, and fill the frame. A crumpled thermal receipt shot at an angle in dim light will produce garbage no OCR engine can fix, so it is worth taking the picture twice.
Making Scanned Documents Searchable
A scanned PDF is a stack of pictures — searching it finds nothing because there is no text layer. Running the pages through OCR gives you text you can search, quote, and archive alongside the scan. For long documents, work page by page; for a whole archive, dedicated batch OCR software will be far less tedious.
Reading Text in Another Language
Photograph a sign, a menu, a form, or a label in a language you do not read, extract the text, then paste it into a translator. Typing unfamiliar scripts — Arabic, Devanagari, Chinese, Japanese, Korean — is impractical for most people, which makes OCR the only realistic route from a photograph to a translation. Select the matching source language before scanning.
Accessibility and Reading Support
Text locked inside an image is invisible to screen readers. Extracting it makes the content available to assistive technology, and pairing OCR with our Text to Speech tool turns a photograph of a page into something you can listen to. This is also a practical way to write accurate alt text for an image that contains text.
Examples
A Screenshot of a Document
A crisp screenshot of body text — the easiest case for OCR.
Screenshot of a paragraph at screen resolution Exact text, typically 95–99% confidence A Photograph of a Printed Page
A phone photo of a book page, shot flat with good light.
Photo, 12 MP, page filling the frame Accurate text with occasional errors on unusual punctuation, 88–95% confidence A Low-Contrast Receipt
A faded thermal receipt — the hard case.
Phone photo of a crumpled receipt in indoor light Partial text, 50–70% confidence — expect to correct the numbers by hand Image to Text (OCR) vs a cloud OCR service
Google Cloud Vision and similar services are more accurate on hard inputs. The trade is your document.
| Feature | Toolorah | a cloud OCR service |
|---|---|---|
| Image leaves your device | Never | Yes — uploaded for processing |
| Accuracy on clean printed text | Very good | Very good |
| Accuracy on hard inputs | Moderate | Substantially better |
| Handwriting | No | Yes on the better services |
| Cost | Free, unlimited | Per-page pricing beyond a free tier |
| Signup or API key | None | Required |
| Works offline | Yes, after the model caches | No |
Tips for Using Image to Text (OCR)
- Higher resolution helps up to a point. Around 300 DPI equivalent is the sweet spot; beyond that you gain little and the scan takes longer.
- Straighten the image before scanning. Tesseract handles small rotations but text at more than a few degrees off horizontal drops accuracy sharply.
- Crop out everything that is not text. Logos, photographs, and heavy borders all give the recogniser noise to misread.
- Match the language before you scan. English mode reading Hindi will produce nothing useful, and a mismatched script is the most common cause of a blank result.
- Always read the output. Even at 95% confidence, roughly one character in twenty may be wrong — and OCR errors in numbers are the dangerous kind.
Frequently Asked Questions
Is my image uploaded anywhere?
No. Tesseract is compiled to WebAssembly and runs inside your browser, so the image is processed on your own device and never transmitted. The only network request is for the language model — a few megabytes of trained data downloaded from a public CDN the first time you use a given language, then cached by your browser. That request contains no part of your image.
How accurate is the recognition?
It depends almost entirely on the input. Clean screenshots and high-quality scans of printed text typically reach 95–99% accuracy. Good phone photographs of printed pages land around 88–95%. Low-contrast, skewed, blurry, or unusually styled text can fall below 70%. The confidence score shown with each result is a reasonable guide: above 80% the output is usually worth trusting after a read-through, below 60% you should treat it as a starting point rather than a transcript.
Can it read handwriting?
Not reliably. Tesseract is trained on printed typefaces, and handwriting — cursive especially — falls well outside what it recognises. Very neat, well-spaced block capitals sometimes work. For anything else you need a model trained specifically on handwriting, such as those in Google Lens, Apple Notes, or Microsoft OneNote.
Which languages are supported?
This tool offers English, Hindi, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Arabic, Simplified Chinese, Japanese, and Korean, plus a combined English + Hindi option for mixed documents. Tesseract itself supports over 100 languages; the list here covers the ones most requested and most reliable. Each language downloads its own model on first use.
Why did it return nothing?
Usually the language is mismatched — an English model reading Devanagari or Arabic finds no recognisable characters. Other causes are an image with no actual text, contrast too low to separate characters from the background, or text rotated more than a few degrees. Try selecting the right language, cropping tighter around the text, and straightening the image.
Why is the first scan slow?
The first run in a session downloads the Tesseract WebAssembly core and the language model for your chosen language — a few megabytes in total. Both are cached by your browser afterwards, so subsequent scans start immediately. Recognition itself typically takes one to five seconds depending on image size and your device.
Can it read PDFs?
Not directly — this tool takes images. For a scanned PDF, export or screenshot the pages as images first, then run them through. If the PDF already contains a text layer (you can select the text in a PDF reader), you do not need OCR at all: copy it directly, or open it in our PDF Editor.
Is there a file size limit?
No fixed limit, but very large images are slow because the whole bitmap is processed in memory on your device. Images above roughly 4000 pixels on the long edge rarely improve accuracy over a well-cropped 2000-pixel version, and they take considerably longer. Cropping to just the text is faster and usually more accurate.