Image to text
Extract English printed text from one image. Check it. Copy it.
Processed on your device
One JPG or PNG · Up to 10 MB and 4 MP · Your image stays on this device.
From image to editable text
Choose or drop one JPG or PNG with clear English printed text.
Up to 10 MB · 4 MP · 4096 pixels per side
From screenshot to usable text
This English OCR tool turns printed letters in a static image into editable plain text. It is useful for a screenshot you cannot select, a clear scan, or a photograph of an English paragraph. Compare the result with the original before using it, especially numbers, punctuation and names. It does not reconstruct a document’s layout.
- Choose one static JPG or PNG image with clear English printed text.
- Select Extract text. On first use, recognition files load from this site; the image itself is processed on your device.
- Check the original, correct any mistakes, and select Copy text or Save .txt.
A checked example
An independent public OCR test image begins “This is a lot of 12 point text to test the”. The local English recognition experiment recovered the marked text without character errors after whitespace normalization. That comparison rule was used only to measure the sample; this tool keeps the recognizer’s text and line breaks. A successful example is not a guarantee for your image: small, blurred or distorted characters can still be wrong. For best results, use a well-lit, upright image rather than a heavily compressed thumbnail.
Which images can I use?
One static JPEG or PNG, up to 10,000,000 bytes (10 MB), 4,000,000 pixels and 4096 pixels on either side. All limits apply together. Files are checked before decoding; animated PNG, WebP, GIF, SVG, HEIC and PDFs are not accepted. The selected image is not silently resized. Transparent areas are shown and recognized on white. EXIF orientation is applied before recognition.
Can it read Chinese, handwriting or tables?
This version supports English printed text. It has no Chinese or multilingual quality guarantee and does not support handwriting, searchable PDF output, table cells or layout reconstruction. A multi-column page may not be read in the order you expect. Check and edit the plain text before relying on it. Unsupported writing may produce no text or incorrect English-looking text; automatic language detection is not provided.
Why does first use take longer, and can I cancel?
Extract text loads the local recognition engine and English model on demand. First use can require several megabytes of site files; later requests may use your browser’s HTTP cache. Preparing recognition is limited to 45 seconds and recognition to 30 seconds, within a shared 60-second task limit. Reading the chosen image has a separate 15-second limit. Cancel stops the task, keeps the validated image and allows a retry. Network speed, device memory and browser background restrictions can affect completion.
Are my text and images uploaded or saved on the site?
No. Images, filenames, recognition output and your edits stay in page memory. Leaving, refreshing or clearing discards them. Copy or save text before changing tools. Optional feedback only sends the rating and message you actively choose to submit; your image or extracted text is never attached. Analytics counts fixed actions and source categories separately. See Privacy for details.
How do Copy text and Save .txt work on phones?
Copy text asks your browser to copy the current edited value. If permission is denied, select the text in the result box and copy it manually. Save .txt starts a UTF-8 plain-text download named extracted-text.txt. Your browser decides its download location; a started download does not prove the phone saved the file. There is no automatic clipboard write or download.