How to extract text from an image
Choose a JPG, PNG or WebP file under 10 MB and 24 megapixels. Select the language that matches the typed text, then start extraction. Keep the page open while it reads. You can cancel and choose a smaller image if the device struggles.
The first run downloads the Tesseract OCR engine and language files from this website. The image itself stays on your device. Later reads can reuse cached engine files. Browser speed depends on your device, image size and the amount of text.
- Crop empty margins and rotate the image upright.
- Choose English, French, or both languages.
- Run extraction and compare the result with the image.
- Correct errors in the editable text, then copy or download TXT.
Improve the image before repeating OCR
Use a sharp original rather than a screenshot of a compressed preview. Avoid glare, skew and shadows. Crop to the relevant text without cutting off characters. Small print benefits from a higher-resolution source, while an enormous image can slow down a phone.
Check O versus 0, I versus 1, accents, decimal separators and dates. This is typed-text OCR, not a guarantee of character-perfect output. Handwriting and complex multi-column tables are difficult cases. This tool returns text; it does not map invoice line items into spreadsheet columns.
When you need invoice fields instead of plain text
Recognizing a word is different from identifying its role. An invoice can have several dates, company names and totals. Copying all the words into a TXT file does not decide which is the supplier, due date or amount due.
The separate invoice workspace adds field mapping, source comparison and structured exports. It has its own access and privacy conditions. The browser text tool is useful for immediate text retrieval, while invoice processing needs a reviewed workflow.