Send result to:
A PNG of a page is usually a screenshot - of a document you cannot copy from, a table in a web page, a form in an app - or a page a scanner saved as an image. Either way it holds no text, only pixels. Paste it into Word and it stays a picture: nothing in it can be selected, corrected or copied. This app rebuilds the page instead. The words and their positions come from recognition, the structure of the page - headings, columns, tables, captions - is worked out separately, and the page is written again as real text, real tables and real pictures in a Word (.docx) file.
PNG keeps every pixel as it was, so letters in a screenshot stay sharp and recognition has a clean page to read. What costs accuracy is size: a screenshot taken of text that was small on screen, or scaled down afterwards, has fewer pixels per letter than a scan. Capture the text at full size when you can.
Because the structure is found before the text is placed, a table in the image comes back as a Word table with its own columns, not as a wall of loose lines, and a form comes back as its captions and values in pairs. The language is worked out from the page itself. Words that recognition is unsure of are read again from the image rather than guessed, and numbers and identifiers are never rewritten. When the file is ready, one line tells you how many pages were rebuilt, how many words were corrected and how many were left in doubt.
A screenshot has no paper size, so you choose the sheet. On A4 or Letter each image is laid whole in the middle of the page, turned to portrait or landscape to match it; "As in the picture" keeps the image's own proportions. Upload several PNG files at once and they become the pages of one document in the order you added them.
There are limits worth knowing. The text is rewritten in one standard font, so the page will not match the image letter for letter: the layout is kept, the typography is not. Handwriting is not what recognition is built for. Read the result before you rely on it, and keep your original.
Nothing to install: the work is done on our servers, with no registration and no CAPTCHA. Up to 10 files at a time, 10 MB each. Uploaded files and download links are removed after 24 hours.
OCR reads the words in the image and where they sit, the structure of the page is worked out separately, and the page is written again as real text, tables and pictures in a .docx.
Files are deleted from our servers after 24 hours, and download links stop working after that.
Drag the PNG files into the white area, or click it to pick them, choose the page size and press CONVERT. When the pages are rebuilt, download the .docx. Up to 10 files at a time, 10 MB each.
Yes - most PNG files are screenshots. Capture the text at full size and do not scale the screenshot down: small, blurred letters are read less reliably. A screenshot has no paper size, so choose A4 or Letter to get an ordinary page, or "As in the picture" to keep its proportions.
Yes. The structure of the page is worked out before the text is placed, so a table comes back as a Word table with its own rows and columns that you can edit or copy into a spreadsheet. Check the numbers before you rely on them.
No, one. The images become the pages of a single document in the order you uploaded them, so a document captured screen by screen comes back as one document.
Files are deleted from our servers after 24 hours, and the download links stop working after that. No one else has access to them.
Yes. Everything runs on our servers, so any modern browser will do - Chrome, Firefox, Safari or Opera, on any operating system.
The layout is kept - headings, columns, tables and pictures stay where they were - but the text is rewritten in one standard font, so it will not match the image letter for letter. Handwriting is not what recognition is built for: check it before you rely on the document.
