Extract printed text from screenshots, scans and photos. Choose a language, review the words beside your image, then copy or download editable text.
Your image stays in your browser. OCR can make mistakes, so always review the result.
Select or drop one JPG, PNG, WebP, BMP or AVIF image. The full-screen workspace opens with your source preview. Use a screenshot, scan or photograph where the letters are readable at their original size.
Choose the language printed in the image. Start with Automatic page layout. Use One block of text for a paragraph, or Scattered text / labels for words separated across a picture. Rotate sideways text until it reads upright.
Select Extract text. The tool loads the OCR engine and the selected language, then reads the image in your browser. First use can take longer because these files need to download. Cancel is available while extraction is running.
Compare the result with the source image. Correct mistakes directly in the text box, then copy the text or download a UTF-8 TXT file. Keep your original image so you can check uncertain words later.
OCR means optical character recognition. A normal photo contains pixels rather than selectable letters. The recognizer looks for character shapes and turns them into text you can edit, search or paste into another application. This tool uses Tesseract.js to run that process on your device.
It is useful when retyping would take time, especially for clear printed paragraphs and screenshots. It does not understand a document the way a person does, and it cannot guarantee that every word, number or line has been recovered. The output box is deliberately editable so you can correct the result before saving.
Recover a paragraph from an app, presentation slide or image when normal text selection is unavailable. Use the original screenshot rather than a compressed copy from a messaging app. Small interface labels may need a closer crop.
Extract printed passages from a handout or document you are allowed to use. Photograph the page straight on with even lighting. OCR can save retyping, but line breaks and the order of separate columns may need correction.
Copy a product name, address or list of items into your own notes. Narrow receipt columns, faded ink and small prices are challenging. Check every amount, decimal point and date against the original before relying on it.
Recover wording from a flattened graphic or exported image when the editable source is unavailable. Decorative lettering and text over a busy background are less reliable than plain printed text with clear contrast.
Choose a format your browser can open. The tool prepares a still raster image for recognition and uses a white background behind transparent pixels. Use still images; animation and multiple pages are not extracted as a sequence.
| Format | Input support | What to know |
|---|---|---|
| JPG / JPEG | Supported | Useful for camera photos and scans. Strong compression can create artifacts around letters. Use the clearest source available. |
| PNG | Supported | A useful choice for screenshots and text graphics because it can preserve sharp edges without lossy compression. |
| WebP | Supported | Works when your browser can decode the image. Recognition depends on the detail retained in the source, including any compression. |
| BMP | Supported | Useful for bitmap scans, although files can be large. Larger files do not automatically give more accurate text. |
| AVIF | Browser dependent | Requires AVIF decoding in your browser. If the image cannot be opened, convert it to PNG or JPG first. |
| HEIC, TIFF, PDF and SVG | Convert first | These are not direct inputs for this tool. Export a page or convert an image to a supported raster format first. This is a single-image OCR tool, not a PDF document processor. |
Need to prepare an unsupported file? Use HEIC to JPG for a phone photo or AVIF to PNG when AVIF cannot be opened in your current workflow. Changing formats will not restore text detail that was already lost.
A tight crop removes unrelated graphics and can give the recognizer a simpler layout. Leave a small border around the letters. For a very large photograph, crop individual sections instead of shrinking an entire page until the characters become tiny.
Use the rotation buttons for text photographed sideways. They turn the image in 90-degree steps. This tool does not correct perspective or flatten curved pages, so retake heavily tilted photos or straighten them before extraction.
Sharp character edges matter more than the file extension or file size. Enlarging a blurry image does not recover missing letters. Avoid repeated JPEG compression, digital zoom and screenshots of screenshots.
The selected language supplies the recognition model and word patterns. Choosing English for another script is likely to produce poor results. This version uses one selected language at a time and does not automatically detect or translate languages.
Automatic layout is a useful first choice for document images. One block can help with a simple paragraph. Scattered text can help with isolated labels, but its reading order may differ from the arrangement you expect.
Check similar characters such as O and 0, I and 1, rn and m. Review names, URLs, email addresses, punctuation and currency amounts. A result that looks plausible can still contain errors.
For preparation, try Crop Image to isolate a paragraph or Adjust Image to improve contrast. Compare any edited version with the original; heavy sharpening or contrast changes can damage thin strokes.
Choose English, Spanish, French, German, Portuguese, Italian, Dutch, Arabic, Hindi, Simplified Chinese, Japanese or Korean. Only the selected language is requested for each extraction. Recognition quality varies with the script, font, image quality and layout. Vertical writing and unusual typography may require manual transcription.
The download is a UTF-8 TXT file containing the text currently in the editor, including your corrections. It has no fonts, embedded pictures, table cells or document styling. You can paste it into a word processor and format it there. The tool does not translate, summarize or verify the meaning of the text.
The tool does not upload your image or recognized text. It downloads the OCR engine from ImgTweak and language data from jsDelivr, then runs recognition locally. These hosts receive normal requests for those software files. The browser may cache engine files and language data for later use; the tool does not intentionally save your image or text in that cache.
Your result is kept in the current editor session. Going back to the page lets you resume it, but refreshing, closing the tab or navigating away can discard it. Copy or download anything you want to keep.
OCR requires more processing than a simple image resize. The engine only starts loading after you select Extract text. Recognition runs in a background worker to keep the interface responsive, but it still uses CPU and memory. A phone may take longer than a desktop, and a busy page may be harder to read than a short label.
There is no fixed MB cap. To reduce processing memory, oversized images are scaled within 12 million pixels and a 6000-pixel longest edge, with a visible notice. Your browser must still decode the original. If it struggles, crop the relevant text before trying again. First use needs internet access, and cached offline operation is not guaranteed.
Technical background: Tesseract.js documentation and Tesseract image quality guidance explain the engine and common recognition problems.
OCR can mistake repeated dots and decorative separators for letter shapes. Ignore dotted separator lines is enabled by default to suppress long, regular rows of small dots on light backgrounds before recognition. It leaves short ellipses alone and reports when it omits a line. Turn it off and extract again if the dots are meaningful. This reduces a common error but does not guarantee that every separator will be detected, especially in tilted or blurry photos.
Yes. You can extract, edit, copy and download text without an account. Processing uses your device, and downloading the OCR engine and language files uses your internet connection.
No. This tool performs recognition in your browser and does not upload the image or extracted text. The OCR engine is served by ImgTweak, and language files are fetched from jsDelivr. Those asset requests do not contain your image.
Do not rely on this tool for handwriting. Its Tesseract engine is intended primarily for printed text. Handwritten notes, joined letters, signatures and decorative fonts can produce incomplete or incorrect results.
The tool offers English, Spanish, French, German, Portuguese, Italian, Dutch, Arabic, Hindi, Simplified Chinese, Japanese and Korean. Select one language per extraction. It does not automatically detect the language or translate the result.
No. The output is editable plain text. Fonts, colors, images and table structure are not preserved. Line breaks may need editing, and text from columns or scattered labels can appear in an unexpected order. It does not export Word, Excel or searchable PDF files.
There is no fixed upload limit in MB. Your browser must still decode the image within the memory available on your device. For OCR processing, images are reduced when needed to fit within 12 million pixels and a longest edge of 6000 pixels. The editor shows a notice before extraction if this applies. Cropping the text area first can preserve more useful detail.
The OCR engine and the chosen language need to load before recognition can start. Later attempts may reuse cached assets, but caching depends on browser settings and available storage. Processing time also depends on your device, image dimensions, language and page complexity.
It works in compatible mobile browsers with WebAssembly and Web Workers, but large images may be slower or exhaust memory. Keep the tab open while it runs. An internet connection is needed for the first engine and language downloads. Offline use is not guaranteed because browsers may clear cached files.
Common causes include tiny letters, blur, poor contrast, a sideways image or the wrong language. Try a clearer source, crop the text area and change the layout setting. An empty result does not prove that the image contains no text.
This version processes one image at a time. Copy or download the current text, then choose Change image for the next file. Keeping one OCR job active helps control memory use on phones and other devices.
Start with a clear screenshot or printed page, extract the text and review it before saving.
Choose an image