Image to text.
OCR that stays on your device.
Pull the words out of screenshots, scans, and photos of documents — the OCR engine runs in your browser, so receipts and IDs never leave your computer.
How do I extract text from an image? Choose a screenshot or scan, pick the language, and press Extract text. The words appear in an editable box — copy them or download a .txt. First run downloads the engine (~8 MB, cached after).
Image to text OCR
Powered by Tesseract compiled to WebAssembly — the same open-source engine behind many desktop OCR tools, running locally. Works best on screenshots and straight, well-lit scans; skewed or blurry photos lose accuracy. The engine (~8 MB) is downloaded once and cached, after which the tool works offline.
How to extract text from an image
- Choose the image — screenshots give the best results, then flat scans, then photos.
- Pick the language — English or Spanish; the model for that language loads once.
- Extract and edit — the text lands in an editable box with a confidence score; fix any misreads and copy it out.
What OCR is good for
- Screenshots — pull a quote, error message, or table out of an image without retyping.
- Receipts and invoices — digitize amounts and line items for expense reports.
- Old scans — make text searchable again.
- Notes and whiteboards — turn a photo of handwriting-adjacent print into editable text.
Good to know
OCR accuracy is mostly about the input image: straight, well-lit, and at least ~300px of width per text column. If a photo comes out skewed, straighten it first. The extracted text keeps line breaks but loses columns and formatting — for PDFs with a real text layer, use PDF to text instead, which is exact. Related: PDF to Word and text to speech.
How accurate is the OCR?
For clean screenshots and scanned documents, accuracy is typically 95%+. Photos of text taken at an angle, in poor light, or with small fonts do worse. Sharpen and straighten the image first — the tool reports a confidence score so you know when to double-check.
Which languages are supported?
English and Spanish are built in. The recognition model for the selected language is downloaded once (a few MB) and cached; more languages can be added to the page over time.
Is my image uploaded?
No. The OCR engine (Tesseract compiled to WebAssembly) runs inside your browser and your image never leaves your device — safe for receipts, IDs, and internal documents.