Translate what you can see, on your device

Two pieces of work happen, and both happen on this device. First the reader finds the text: a PDF that carries real text gives its words up directly, and a scanned page, a photograph or a manga page is read by the same neural reader the OCR tools use — about sixteen megabytes the first time, cached by your browser afterwards. Then your browser's own translator turns each block into the language you chose. The first time you translate a pair of languages, the browser downloads that pair's language pack from its own vendor. That is a request the browser makes, not this page, and it carries nothing of your file. There is no translation server of ours, because there is no server here at all.

Today each tool gives back the text it read, block by block, with the translation beside it. Both sides are editable, because the reading is a guess and the translation is a second one: correct what was read and that block can be translated again on the spot. The translation can be copied, or downloaded as plain text or Markdown with the page breaks or the bubble numbers kept. Setting the translation back into the picture, re-lettering a manga page and a translated PDF with its layout kept are not built. Each tool has a button for them that says so on the next screen: nothing is charged, no account is asked for, and the press is counted as a vote for what to build first.

The reading works in every browser this site supports. The translating needs the browser's built-in translator, which is in Chrome 138 or Edge 148 and newer, on a laptop or desktop. In Safari, on a phone or in an older browser, each tool says so before you choose a file, still reads the text out of it, and leaves the text ready to copy into any translator you trust.