Extract editable text from screenshots, receipts and photos β 12 languages, fully local
Add a screenshot or photo β recognition happens on your device
Recognition runs locally with Tesseract.js. The selected language model (15β20MB) downloads once and is cached. Printed text is recognised reliably; handwriting, low-contrast photos and stylised fonts are not β proofread anything you plan to submit.
Tesseract is a printed-text engine. It does well on screenshots, scanned documents, receipts, and slides β anything with consistent letterforms and decent contrast. It struggles with handwriting, decorative fonts, text over photos, and very low-resolution crops. If a result looks garbled, upscale the image 2Γ first and run it again; Tesseract is sensitive to character height.
Everything happens on your device. That matters more than it sounds: the images people most want to extract text from β invoices, bank statements, ID cards, chat screenshots β are exactly the ones they should not be uploading to a stranger's server. The only network request is the one-time language model download.
Extract text from images free with browser-based OCR. Paste a screenshot or drop up to 10 photos, choose from 12 languages, and copy or download editable text. No upload.
OCR does not βreadβ an image the way a person does. It binarises the image, finds connected components that look like glyphs, segments lines and words, and matches shapes against a language model. That pipeline explains both its strengths and its failure modes: clean printed text has consistent glyph shapes, while handwriting, shadows and perspective distortion break the segmentation before recognition even starts.
Three inputs dominate quality: resolution, contrast and crop. Aim for at least 20 pixels of character height, avoid JPEG mush around small text, and crop out irrelevant margins so the layout analyser is not confused by columns. If a result is close but wrong on names and numbers, proofread those fields manually β OCR errors cluster on similar-looking characters (0/O, 1/l/I, 5/S), which is exactly where a silent mistake is most costly.
Each language is a separate trained model, not a setting inside one universal engine. That is why the first run downloads 15β20MB and later runs start instantly: the model is cached by the browser. It also explains an honest limitation β you pick one primary language per run. A page that mixes Chinese and English will recognise the dominant script better than the other, and switching languages mid-document is not supported. For mixed documents, run the page twice with different languages and merge the results.
On clean, high-contrast printed text β screenshots, scans, slides, receipts β accuracy is high enough to use with light proofreading. It degrades quickly with small text, low contrast, perspective skew, decorative fonts and handwriting. A useful trick: upscale a blurry crop 2Γ before running it, because Tesseract is sensitive to character height.
Twelve: English, Simplified and Traditional Chinese, Spanish, French, German, Portuguese, Japanese, Korean, Arabic, Russian and Hindi. Each language is a separate 15β20MB model downloaded on first use and cached afterwards. Mixed-language documents work best when you pick the dominant language; switching mid-document is not supported.
Poorly, and the page says so rather than overpromising. Tesseract is a printed-text engine; handwriting has connected strokes and inconsistent letterforms it was not trained on. For handwritten notes, use the dedicated handwriting tool, which uses a different model.
Yes β press Ctrl+V (β+V on Mac) anywhere on the page and the clipboard image is added to the queue. This is the fastest path for chat screenshots and error messages, and it avoids saving a file just to upload it again.
No. The OCR engine and language model run in your browser; only the model file is downloaded. Receipts, bank statements and ID screenshots stay on your device, which is the point β those are exactly the images you should not hand to a free OCR website.
What do you call a cat that loves the beach?
No signups, no data sold. Every tool is free to use β the free tier allows 5 downloads or saves per day, and the optional Pro plan ($7/mo, $59/yr) adds unlimited downloads, batch processing, white-label exports and an ad-free experience.
βSupport me on Ko-fiβ support the free tier100% of proceeds go towards hosting & building more free tools.