Text from photos, screenshots and scans
Sooner or later everyone ends up retyping something that already exists as a picture: a paragraph from a textbook page, an address in a screenshot, a quote on a slide, the conditions printed on a receipt. Image to text reads the printed characters in the picture with OCR (optical character recognition) and hands them back as real text that you can select, copy and edit.
It accepts JPG, PNG, WEBP, GIF and BMP images, plus HEIC photos from an iPhone when your browser can open them, as Safari does. You can add several images in one go, which helps when a document was photographed page by page: each image is recognized in turn.
Pick the language of the text
Recognition works in English, Spanish and Portuguese, and you can tick more than one when a page mixes them, for example an English form filled in with Spanish names. The choice matters: it is how the tool knows that an ñ or an ã is a letter and not a smudge on the paper.
The first time you use a language, its recognition data is downloaded from ConvertirDocs. It is a few MB, so the first run takes a little longer on a slow connection. What travels is that language data, never your image.
Copy it, fix it, or download TXT or Word
The recognized text appears in an editable box. Fix a wrong character, delete the lines you don't need, then copy everything and paste it into an email, a chat, a spreadsheet or your notes. You can also download it as a plain .txt file or as a Word document (.docx).
The Word file contains the text in paragraphs and nothing else. It doesn't rebuild tables, columns, fonts or the position of things on the page. If you need a file that looks like the original, keep the image; if you need the words, this is the quick way to get them.
What OCR reads well, and what it doesn't
OCR does well with clear printed text: book pages, letters, invoices, signs, screenshots of websites and apps. It does not read handwriting well, and it struggles with blurry photos, pages shot at a steep angle and text with little contrast against the background. It doesn't reconstruct tables either: the cells come out as loose lines of text.
A few habits improve the result. Photograph the page straight on and in good light, fill the frame with the text, and cut away everything else with “Crop image” before recognizing. When numbers matter, such as amounts, dates or reference codes, always check the result against the original.
Your images stay on your device
Recognition runs inside your browser, on your own phone or computer. The picture isn't uploaded to a server and nothing is kept once you close the page, which matters when the image is a bank statement, a medical note or an ID card. It's free, with no account, no email and no limit on how many images you convert.