Image to Text OCR
Extract editable text from screenshots, receipts, and photos
The OCR engine and language data load on demand; recognition runs locally in your browser. First use may take a few seconds.
Editable OCR text will appear here.
About this tool
Extract editable text from printed screenshots, receipts, labels and photographed documents using a caller-owned Tesseract worker in your browser. Choose a PNG, JPEG, WebP or BMP image up to 25 MB, then English, Simplified Chinese or both and a page layout. OCR engine and language data download on demand; recognition itself runs locally. Small images may be upscaled, which cannot restore missing detail. Edit, copy or download TXT/JSON after recognition. The confidence value describes the original engine result and is not recalculated after edits. JSON retains the recognized file, language and layout with an edited flag. Handwriting, blur, glare and complex tables need human review; the output is text, not a spreadsheet model. Inputs are not saved as drafts.
Scenario Recipes
Capture text from a receipt or screenshot
Goal: Extract a reusable draft without treating OCR output as an automatically verified record.
- Use a sharp, tightly cropped image with minimal glare and choose the matching language model.
- Run recognition and compare totals, names, dates, and line breaks with the source image.
- Edit the result before copying it or exporting TXT or JSON.
Result: Manual retyping is reduced while important fields still receive a deliberate accuracy check.
Frequently Asked Questions
What images work best?
High-resolution, well-lit images with dark text on a clean background generally produce the best results. Crop away unrelated borders and glare when possible.
Can it read handwriting?
It is optimized for printed text. Handwriting recognition is not guaranteed and usually needs manual correction.
Can I edit the OCR result?
Yes. The result is shown in an editable text area before you copy or download it.
Does it support Chinese?
Yes. Select Simplified Chinese or the combined English plus Simplified Chinese model.
Is the image sent to a server?
No. Recognition runs locally after the browser downloads the OCR engine and language data.
Why are table columns not preserved?
OCR returns text order and confidence, not a true spreadsheet model. Exported text should be checked before using it as structured data.
What does confidence mean after I edit the result?
It remains the original OCR engine estimate, not an accuracy score for your edits. JSON identifies it as original_ocr and marks edited text. Changing the source or OCR options clears the result.
Keep browsing