Image to Text (OCR)
Turn a photo, screenshot or scan into text you can copy — on your device, in 13 languages.
Recognise text
Images
Opening the image…
To read only part of the image, drag over it (on a phone, select Select area first). Keyboard: focus the image and press Enter, then the arrow keys to move the area and Shift + arrow keys to resize it.
Only this language’s recognition data is downloaded.
Documents, pages and screenshots with several blocks of text.
The first run downloads the OCR engine and the English data (about 7 MB) from this site; later runs use your browser’s cache. Your images stay on this device.
Recognised text
Words to check
Select a word to see it on the image.
OCR reads printed text. Handwriting, tables, decorative fonts and blurry photos are unreliable — always check names, numbers and amounts against the original.
About the Image to Text (OCR)
Copy the words out of a photo, a screenshot or a scanned page instead of typing them again. Choose one or more images, pick the language, and the text appears in a box you can edit, copy or save as a TXT file. Words the engine is unsure about are outlined on the image so you know what to check.
Recognition runs on your device with Tesseract 5, the open-source OCR engine, and its trained data for English, Hindi, Marathi, Bengali, Gujarati, Tamil, Telugu, Kannada, Malayalam, Urdu, Arabic and Chinese (Simplified and Traditional). Nothing is uploaded. You can read just part of an image by drawing a box around it, straighten slightly tilted text, and clean up photos taken in uneven light.
How to use it
- Choose, drop or paste one or more images (JPG, PNG, WebP, HEIC, GIF or BMP). Screenshots can be pasted straight from the clipboard with Ctrl+V (⌘V on a Mac).
- Choose the language of the text. For a Hindi, Tamil or other Indian-language document that also has English words, tick Also recognise English.
- Optional: drag a box on the image to read only that area, rotate a sideways photo, or pick a layout that fits the picture (for example One line for a serial number or Scattered text for a poster).
- Select Recognise text. Check the words outlined on the image, correct the text if needed, then Copy it or Download TXT. With several images, Download all saves one file with every image’s text under its name.
Examples
Paste the screenshot · English · Automatic layout
Editable text; dark-mode screenshots are inverted automatically for better accuracy
Hindi + Also recognise English · Clean up photo on
Devanagari text with the English words kept, low-confidence words outlined for checking
Draw a box round the number · One line layout
Just the line you selected, e.g. “SN 4821-77XQ-0093”
Common uses
- Copying text from screenshots of chats, emails, slides and error messages.
- Getting the text of a printed letter, notice or form from a phone photo.
- Reading serial numbers, order numbers and addresses off labels.
- Extracting quotes from photos of book pages for notes and study.
- Turning Hindi, Marathi, Tamil and other Indian-language printouts into editable text.
Getting the best results
- Use a sharp, well-lit picture taken straight on. Text should be at least about 20 pixels tall; small screenshots are enlarged automatically before reading.
- Crop or select the area with text: logos, photos and table lines around the text cause stray characters.
- Pick the right language. Recognising English text with the Hindi model (or the other way round) gives poor results.
- Clean up photo evens out shadows and boosts contrast for photographed pages; turn it off for clean screenshots if a light-coloured text disappears.
- Straighten tilted text corrects small tilts of a few degrees. Rotate sideways or upside-down photos with the rotate buttons first.
Layouts
- Automatic: finds blocks, columns and paragraphs — best for documents and most screenshots.
- One column: a single column of text in different sizes, such as a receipt or an article.
- One block: one uniform paragraph.
- One line: a heading, sign, number plate or serial number.
- Scattered text: words spread across a picture, such as posters, memes and app screens.
Keep extra spaces preserves runs of spaces between words, which helps with simple aligned tables (the result is still plain text, not a spreadsheet).
What OCR cannot do well
Tesseract is trained on printed text. Handwriting is usually not recognised, and tables come out as lines of text without their columns and borders. Decorative or very stylised fonts, text on busy backgrounds, curved text, very small or blurry text and low-resolution photos all reduce accuracy. Always check names, numbers and amounts against the original before you rely on them.
Limitations
- Printed text only: handwriting, signatures and CAPTCHAs are not recognised reliably.
- Tables are returned as plain lines of text, without columns.
- Thirteen recognition languages are available; Punjabi, Odia and other languages are not offered yet.
- Very large photos are scaled to about 24 megapixels for recognition, to fit in phone memory.
- Up to 40 images per batch; they are read one after another.
Privacy
Your images are read in your browser and never sent to a server. The first time, the OCR engine and the chosen language’s recognition data (about 7 MB for English) are downloaded from MySmartCoPilot, not from a third party, and your browser keeps a cached copy; nothing about your images is stored.
Frequently asked questions
Are my images uploaded to a server?
No. The OCR engine runs in your browser, so your images and the text never leave your device. Only the engine and the language data are downloaded from MySmartCoPilot the first time.
Which languages can it read?
English, Hindi, Marathi, Bengali, Gujarati, Tamil, Telugu, Kannada, Malayalam, Urdu, Arabic, Chinese (Simplified) and Chinese (Traditional). For an Indian-language document with English words in it, choose the language and tick Also recognise English.
Can it read handwriting?
Not reliably. Tesseract is trained on printed text, so handwritten notes usually come out garbled. Neat block capitals sometimes work; always check the result.
Why are some words outlined on the image?
Those are words the engine recognised with low confidence (below 60%). They are the ones most likely to contain mistakes, so check them against the picture.
How do I copy text from only part of an image?
Drag a box over the part you want before selecting Recognise text. Only that area is read. Select Clear area to read the whole image again.
Can I convert a scanned PDF?
Use OCR PDF for PDFs: it reads every page and can save a searchable PDF as well as the text.