Picture Translator: Translate a Picture or Photo to English

In AI mode, which is the default, a reduced copy of the picture (at most 1024 pixels on its longest side) is sent to an AI model that reads and translates it; it is not stored. In Fast mode, no: the picture is read inside your browser and only the text it finds is sent to be translated. You can switch modes above the queue.

Drop the picture in, set "Translate into" to English, and press Start. AI mode works out the language in the picture by itself, reads the text, translates it and draws it back over the picture. Download the result as a PNG at the same pixel size as the original, with no watermark and no sign-up, or copy the original and the English from the list under it.

Yes, if the text in the photo is sharp and reasonably large: a sign, a menu on a table, a page from a book. AI mode copes with angled text and busy backgrounds much better than Fast mode. Handwriting, very low resolution and heavy shadow can still come out wrong or not at all. In Fast mode, when nothing readable is found, the tool says so and stops: nothing is sent and none of your daily allowance is used.

Yes. It runs in your phone's browser, so there is no app to install. Take the photo of the menu, sign or label first, then open this page, tap the drop area and choose the photo; on many phones the picker also lets you take a new photo there and then. Press Start and AI mode reads it, which copes with angled or uneven text in a photo better than Fast mode. Download the PNG to your phone, or copy the translated lines from the list under it.

Yes. Google Translate on the web has an Images tab: you upload or paste a picture, and Google's help page says you can copy the translated text or download the translated image. You get the same two things here, and AI mode is good at comics and pictures where the text sits in bubbles or over artwork. If you would rather not upload the picture at all, Fast mode reads it inside your browser and sends only the words it finds.

Japanese, Korean, Chinese, Arabic, Thai, Hindi, the European languages and many more, including a picture that mixes two of them. AI mode works out the language by itself. In Fast mode, leave "Text in the image is" on Auto-detect or pick the language yourself; the wrong pick comes back as nonsense, which you will see in the list under the result.

60 languages: English, Simplified Chinese, Traditional Chinese, Japanese, Korean, Spanish, Portuguese, French, German, Italian, Russian, Arabic, Hindi, Indonesian, Thai, Vietnamese, Turkish, Dutch, Polish, Ukrainian, Czech, Slovak, Romanian, Hungarian, Greek, Bulgarian, Serbian, Croatian, Slovenian, Swedish, Danish, Norwegian Bokmål, Finnish, Estonian, Latvian, Lithuanian, Catalan, Galician, Hebrew, Persian, Urdu, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi (Gurmukhi script), Nepali, Malay, Filipino, Swahili, Afrikaans, Azerbaijani, Kazakh, Uzbek (Latin script), Georgian, Armenian, Mongolian (Cyrillic script). Pick one from the "Translate into" list on the tool.

Yes, this one in AI mode. Drop in one manga, manhwa or comic page at a time: the text in each speech bubble is translated and drawn back inside the same bubble, and sound effects are translated too. Each page uses 2 of the 5 free pages a day, so two comic pages a day are free; for a whole chapter, the $9 pass covers 1,500 pages for 30 days. Japanese, Korean and Chinese are read without choosing the language first.

The original text is covered with the average colour found around it, and the translation is drawn on top in a plain system font. On a white document or a flat screenshot the patch is invisible; on a photo, a gradient or a busy background you will see it. Fonts, colours, bold, italics and any effect on the original text are not reproduced.

Usually. A translation is often longer than its source, so the type is shrunk to fit its box. When even the smallest size does not fit, the text runs past the edge of the box rather than being cut off.