OCR (Image to Text)
Extract text from images using OCR
Your files are processed on your device. The first use downloads a processing engine, which your browser then keeps.
Drop an image with textTap to choose an image with text
or click to browseor tap to choose
Extract text from images using OCR. You can also paste a screenshot with Ctrl/Cmd+V.
Large files (over 50 MB) may take longer or run out of browser memory
About OCR (Image to Text)
Extracts readable text from an image or scanned document using Optical Character Recognition, turning a picture of text into text you can actually select, copy, and edit. Runs entirely in your browser — the image never leaves your device.
Upload or paste (Ctrl/Cmd+V) an image containing text (JPG, PNG, WebP or an iPhone HEIC photo), choose the language of the text if it is not English, and click Extract Text. Copy the result or download it as a .txt file.
Clear, high-contrast, well-lit scans produce far more accurate results than a blurry photo or a low-resolution image — if accuracy matters, it's worth re-scanning a source document rather than working from a rough phone photo. This reads images specifically, not PDFs — for a scanned PDF, use the PDF OCR tool instead, which handles multi-page documents properly.
Watch: OCR (Image to Text) in Action
- Can this extract text from a scanned PDF?
- This tool reads images specifically, not PDFs — for a scanned PDF, use the PDF OCR tool instead, which handles multi-page documents properly.
- Why does the first run need an internet connection?
- The recognition engine downloads the language data for your chosen language when it is first needed. If that download cannot happen, the tool tells you it could not start and asks you to check your connection. The image itself stays on your device throughout.
- Can I use it on a PDF?
- No. This tool accepts image files only, and a dropped PDF is turned away with a message pointing to OCR PDF. Use that tool for scanned documents, or export a single page of the PDF as an image and drop that here.
- What does the low confidence warning mean?
- If the engine's overall confidence is under 70 percent, a note shows the figure and suggests a clearer image or the right language. It is an average for the whole picture, so some lines may still be fine, but check the text before you rely on it.
- What does Keep original line breaks do?
- By default the tool joins lines into paragraphs and splits at list items or larger gaps. Ticking the box swaps in the raw reading, which keeps each line where it appeared. That works better for addresses, menus, short lists and table rows.
- My phone photo is sideways. Will it still work?
- Usually. Before reading, the tool applies the orientation stored in the photo, and if the first result is weak it also tries 90, 180 and 270 degree turns and keeps the best one. A photo with no usable orientation data and very little text can still fail.
How to use OCR (Image to Text)
Drop an image onto the page, or press Ctrl or Cmd plus V to paste a screenshot straight from your clipboard. JPG, PNG, WebP and iPhone HEIC photos are accepted, and HEIC is converted to JPEG first. A small preview appears. Beneath it, open Language of the text and choose English, English + Hindi, Hindi, Spanish, French, German, Portuguese or Italian. English is selected by default. Press Extract Text and watch the percentage.
When reading finishes, the words appear in an Extracted Text box. Tick Keep original line breaks if the picture holds a list, table or address, then use Copy or Download .txt. The text file takes the image's name. Reset clears the image so you can start with another one.
When to use it
Use it to lift words out of a screenshot of an error message, a photographed receipt, a page from a printed book, a slide photographed at a lecture, or a business card, whenever retyping would take longer than checking a quick transcription. It suits a single image of printed text in one of the eight language choices when you want something you can paste into a note or search.
When not to use it
Skip it for handwriting, decorative logos or text laid over a busy photograph, where recognition is unreliable. It reads one image at a time and does not open PDFs; for a scanned PDF with several pages, the OCR PDF tool on this site is the better fit. If you need a table in a spreadsheet, expect to rearrange the output by hand.
Tips for best results
Crop the picture down to the text before you drop it in, since borders, shadows and background patterns are a common source of stray characters. If the output is gibberish, check the language selector first: reading French with English selected loses accents. Compare the default result with Keep original line breaks, which keeps each printed line instead of joining lines into paragraphs. Treat the amber confidence warning as a cue to proofread.
Limitations
Accuracy depends on the picture. Small, blurred, skewed or low-contrast text will contain mistakes, so proofread anything important. Only the eight language choices listed are offered, and one image is processed per run. Columns, tables and fonts are not preserved; the result is plain text. If the first reading looks poor, the tool reads again as one block of text and tries the image turned by 90, 270 and 180 degrees.
Privacy
The image is read by a recognition engine running in your browser, and the text is produced on your device. Your picture is not uploaded. The engine does need the language data for the language you picked, which is downloaded the first time that language is used; that request carries none of your image. Copy and Download .txt use only your clipboard and a local file.
More from File & Misc: