How to get the best OCR results
Optical character recognition works by comparing the shapes on your image against trained letterforms, so the cleaner the input, the more accurate the output. Aim for good contrast between the text and its background — dark text on a light, even surface reads far better than text over a busy photo or a shadowed page. Try to photograph or scan straight-on rather than at an angle, since skewed or heavily perspective-warped text confuses the recognizer's line detection. A resolution around 300 DPI (or a photo where the text is sharp and legible when you zoom in) gives the engine enough detail to work with — text that's blurry or tiny in the source image will stay unreliable no matter what settings you apply. Cropping out everything except the text block before you OCR it also helps, since stray graphics, borders and other page furniture can be mistaken for characters.
What the confidence colours mean
After recognition, each line is shown with a confidence score reported directly by the OCR engine — how sure it is about that line's characters. Lines below 60% are highlighted so you know exactly what to double-check by eye before you trust the output, lines between 60% and 80% are flagged as medium confidence, and anything at 80% or above is generally safe to accept as-is. This is far more useful than a single overall accuracy number, because a photo often has one blurry corner or one line in a decorative font that drags down an otherwise clean scan — the per-line view tells you precisely where to look.
Why this runs on your device
Most "free" online OCR tools upload your image to a server, which means a photo of your passport, a signed contract, a prescription or a handwritten note passes through someone else's infrastructure before you get your text back. OCR Scan uses Tesseract.js, a WebAssembly build of the open-source Tesseract OCR engine, which runs the entire recognition pipeline inside your browser tab. The only network activity is a one-time download of the engine core and the language data you choose — after that, everything, including every image you process, stays on your device. You can verify this yourself: turn off Wi-Fi after the first run and OCR Scan keeps working.
Languages
Tesseract supports well over 100 trained languages. This tool's picker curates the 20 most requested for everyday use — European languages, several South and East Asian scripts, Arabic, and both Simplified and Traditional Chinese — shown with their native name alongside the English name so they're easy to find. The free tier reads English plus any one additional language from that list at a time; Pro lets you combine several languages in a single recognition pass, which matters for documents that genuinely mix scripts, like a bilingual menu or an international form.
Batch, searchable PDF and tables (Pro)
Scanning a single receipt is a one-image job, but digitising a report, a contract or a stack of receipts is not. Pro's batch queue accepts up to 50 images at once and recognises them in sequence, joining the results into a single document with clear page separators. The searchable PDF export goes a step further: it keeps your original scanned image exactly as it looked, but lays an invisible, precisely-positioned text layer underneath each recognised word, so the PDF looks like a scan but can be searched, selected and copied from like a real document — the same technique professional scanning software uses. For forms and tables, the best-effort table-to-CSV export groups words into cells based on their spacing and hands you a spreadsheet-ready file, saving the retyping.
Free vs Pro
Free covers genuine one-off jobs completely: drop a photo or screenshot, pick a language, clean it up with the pre-processing toggles, and get editable text with confidence highlighting you can copy or download as .txt — no limit on how many times you use it, just one image loaded at a time. Pro (a one-time $8 purchase, not a subscription) is for anyone processing more than a page at a time: students digitising handouts, small businesses archiving paper receipts, researchers extracting text from scanned documents, or anyone who needs a searchable PDF instead of a static image.
Related free tools: PDFLocal Background Remover Squeeze Image — or browse the AppVitamins store.
Frequently asked questions
Is my image uploaded anywhere?
No. OCR Scan runs the Tesseract.js OCR engine entirely inside your browser using WebAssembly. Your photo, screenshot or scan never leaves your device — the only network request is downloading the language pack once, which is then cached for offline use.
Does it read handwriting?
Printed text — books, receipts, screenshots, signs, forms — works best. Neat, well-separated handwriting can work but is hit-and-miss, and cursive handwriting is generally not recognised reliably. Tesseract is trained primarily on printed fonts.
Which languages are supported?
The Tesseract engine itself supports over 100 languages. This tool's picker curates the 20 most requested — English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Turkish, Vietnamese, Indonesian, Hindi, Arabic, Japanese, Korean, Chinese (Simplified and Traditional) and Thai. The free tier reads English plus one more; Pro lets you combine several at once.
Why is the first run slower?
The first recognition downloads the OCR engine core and the chosen language's trained data (a few megabytes each) from a CDN. Your browser caches both afterward, so every run after that starts almost instantly and works offline.
How do I OCR a PDF?
Export the PDF's pages as images first (most PDF viewers can "export as image", or screenshot each page), then drop those images in here. Or use PDFLocal to work with a PDF's pages first.