OCR — Image to Text
Use this free online OCR (Optical Character Recognition) tool to extract editable text from any photo, scanned document or screenshot. It is useful for digitising printed receipts, copying text from a non-searchable PDF screenshot, transcribing whiteboard photos after a meeting, capturing serial numbers from a product label, and turning a textbook chapter into copy-pasteable notes. Drop the image into the upload zone, pick the language, and the tool returns the recognised text ready to copy. Recognition runs in your browser using a WebAssembly build of Tesseract — your photos never leave your device, even sensitive ones like contracts, statements or ID cards.
Click to choose or drag & drop an image
About OCR
This tool uses Tesseract.js running entirely in your browser — your images never leave your device. The first run downloads the language model (a few MB), which is then cached. Best results come from clear, high-contrast images with horizontal text.
How to Use This OCR — Image to Text
Converting an image to text takes a drop, a language pick and a click.
- Drop your image into the upload zone or click to choose a file.
- Pick the primary language of the text (English by default; multi-language OCR is supported).
- Click Recognise Text and wait while the WebAssembly OCR engine loads on first use (~10 MB cached after).
- Read the extracted text in the output area.
- Click Copy to send the text to your clipboard, or use the export button to save it as a TXT file.
How This OCR — Image to Text Works
Optical Character Recognition is the process of converting images of text into machine-readable text. The tool uses Tesseract, an open-source OCR engine originally developed by HP and now maintained by Google, compiled to WebAssembly so it runs entirely in the browser. When you drop an image, the engine first runs a layout analysis pass to find blocks, lines and individual character regions, then it normalises each character region (binarisation, deskewing, line straightening) and feeds it through a recurrent neural network trained on millions of text samples per language. The network returns a probability distribution over characters, and a language-model post-processing pass picks the most likely word given the surrounding context. Accuracy is highest on clean, high-contrast print at 300 DPI or above; it drops on handwritten text, low-resolution photos, unusual fonts and complex layouts with multiple columns. Multi-language packs cover dozens of scripts including Latin, Cyrillic, Greek, Arabic, Chinese, Japanese and Korean. Because everything runs locally there is no upload, no character limit, and no privacy concern — sensitive documents like contracts, payslips and ID cards can be OCR'd safely.
Last updated: May 2026.
Frequently Asked Questions
What languages does the OCR support?
The dropdown lists 11 commonly used languages including English, Spanish, French, German, Russian, Arabic, Chinese, Hindi, and Japanese. Tesseract supports 100+ in total.
Why is the extracted text inaccurate?
OCR struggles with low-resolution, skewed, blurry, or stylised text. Try a higher-quality scan, rotate the image to be horizontal, or increase contrast.
Does this work on handwriting?
Tesseract's default models are trained on printed text. Handwriting recognition is possible but requires specialised models — accuracy will be poor with this tool.
How long does OCR take?
A typical screenshot takes a few seconds. The first run also downloads the ~10MB language model. Subsequent runs are fast because the model is cached.
Can I extract text from a PDF?
Use our PDF-to-Word tool for text-based PDFs. For scanned PDFs, convert pages to images first, then OCR each image.