Web toolsImage text extraction (OCR)
Extract text from images
On-device text recognition
Recognize printed text in PNG, JPG, WebP and BMP images on your device. Choose one of 23 recognition languages, then copy the result or save UTF-8 TXT.
Open a receipt photo, choose its text language and compare the recognized text with the original.
Images, file names and recognized text are not sent to servers, analytics or advertising services. The engine and selected language model download from this site; site analytics and ads may load separately.
Choose an image to recognize its text here.
Source image
One PNG, JPG, WebP or BMP, up to 20 MiB, 4,096 px per edge and 4,194,304 pixels total. PDF input is not supported.
Choose one main language. The selected model (about 0.5–6.1 MB) and engine download when needed. Your browser may cache these files.
Recognition result
Some text may be missing or incorrect. Compare with the original before copying or saving.
How to use
- Choose an image containing printed text.
- Select its main language and press “Recognize text”. An internet connection is needed to download the model.
- Check the result against the original, then copy it or save UTF-8 TXT.
Recognition scope and notes
- Works best with clear, upright printed text. Handwriting, complex tables, tilted or blurred text may be recognized poorly.
- Characters, paragraphs and reading order can change. Original formatting is not restored and accuracy is not guaranteed.
- Animations use one still frame and transparent areas become white. Originals are unchanged. PDF files cannot be used as input.
- Jobs stop after 3 minutes and may fail due to device memory limits. You can cancel; closing or reloading clears the image and result.
Use images you created or are permitted to use. Extracting text does not grant permission to publish, distribute or commercially use its contents.
