Extract editable, copyable text from photos, screenshots and scans.
⚡ Powered by the open-source Tesseract.js OCR engine, loaded from a CDN. The model downloads once on first use and is then cached by your browser; your photo itself is never uploaded anywhere.
Text is recognised locally using the Tesseract.js OCR engine (English)
The Image to Text tool reads the words inside a photo, screenshot or scanned document and turns them into ordinary, copyable text using Optical Character Recognition, or OCR. Instead of retyping a paragraph from a printed page or a screenshot by hand, you can upload the image here and receive editable text within seconds, directly inside your browser.
This tool loads Tesseract.js, a widely used open-source OCR engine, from a public CDN the first time you use it, and then runs the actual recognition entirely inside your browser using WebAssembly. The engine analyses the shapes of characters in your image, compares them against trained language models, and reconstructs the most likely sequence of letters, numbers and punctuation, which it then returns to you as plain text you can copy or download.
Students photograph a printed textbook page and extract the paragraph to quote it in an essay without retyping. Office workers pull the text out of a scanned invoice or old document that only exists as an image. Researchers digitise handwritten or printed notes from meetings, whiteboards or books into a searchable, editable format that can be pasted into a document or spreadsheet.
OCR accuracy is highest on clear, well-lit, high-contrast images with plain fonts, so a straight, well-focused photo of a printed page will usually produce far more accurate text than a blurry or angled one. Handwriting, decorative fonts and very low resolution images are more likely to produce mistakes, so always proofread the result before using it somewhere important. Cropping out unrelated parts of the image first with the Image Cropper tool can also improve accuracy by giving the engine only the text you care about.
The engine primarily uses an English language model by default in this tool, though the underlying Tesseract.js engine does support many other languages in general.
No, the recognition itself runs locally in your browser using WebAssembly; only the OCR engine's program code is fetched once from a public CDN, never your photo.
OCR engines work by pattern-matching character shapes, so blurry photos, unusual fonts, low contrast or skewed angles can all lead to occasional misreadings that are worth proofreading.
It can sometimes read very neat handwriting, but OCR engines like this one are trained mainly on printed text and are generally far more reliable on typed or printed documents than handwriting.
