Extract text from any image - privately
Need the text out of a screenshot, a photo of a document, a receipt or a slide? This tool reads it for you and gives you clean, editable text you can copy or download. Unlike most "image to text" sites, your picture is never uploaded - the optical character recognition (OCR) runs entirely inside your browser, so even confidential scans and ID documents stay on your machine.
How in-browser OCR works
OpenConvert uses Tesseract, a mature open-source OCR engine compiled to WebAssembly. The first time you pick a language, a small recognition model downloads to your browser and is cached for next time - the same way the Background Remover loads its model. After that, your images are processed locally with no round-trip to a server.
Tips for the best accuracy
- Use a sharp, high-contrast image - dark text on a light background reads best.
- Straighten it first - heavily skewed or rotated text lowers accuracy.
- Pick the right language before extracting, especially for accented or non-Latin scripts.
- Crop out clutter so the engine focuses on the text you actually want.
What OCR reads well - and what defeats it
Realistic expectations help here. Printed text - screenshots, book pages, receipts, signs photographed straight-on - typically extracts with high accuracy when the text is sharp and at least roughly 20 pixels tall. What degrades results, in rough order of severity: handwriting (largely unsupported by this engine), blur and low resolution, text photographed at an angle or curved across a page, decorative fonts, and colored text on low-contrast backgrounds. Two fixes cost nothing: retake or re-screenshot at higher resolution, and crop to just the text region with the Crop Image tool so the engine is not distracted by surrounding graphics.
Practical uses people put this to
Pulling a quote out of a screenshot instead of retyping it, digitizing a printed recipe or contract clause, extracting a table of numbers from a photographed page, recovering text from an old scan where the original document is gone, and converting whiteboard photos into editable notes. Because the result lands in an editable text box per image, you can fix the occasional misread character before copying or downloading - always proofread numbers in particular, since a misread 8/0 or 1/7 is the most common OCR error class.
FAQ
Are my images uploaded to a server?
- No - the OCR runs locally in your browser using WebAssembly. Your image never leaves your device; only the language model is downloaded to your browser.
Which languages can it read?
- More than 15, including English, French, German, Spanish, Portuguese, Italian, Romanian, Russian, Chinese, Japanese and Arabic. Choose the language before extracting for best accuracy.
Why is the first scan slow?
- The first time you use a language, its recognition model (a few megabytes) is downloaded to your browser and cached. After that, scans are much faster.
How accurate is the OCR?
- It is very accurate on clear, high-contrast printed text. Handwriting, low resolution, glare and skewed angles reduce accuracy - use a sharp, well-lit, straight image for best results.