Image Translator

Translate text in any image in Chrome — manga, memes, screenshots, product photos. Free, no account, nothing uploaded.

🖼️ Image Translator is a free Chrome extension that translates the text in any picture you see in your browser — manga pages, memes, screenshots, product photos, signs, charts and scanned documents — and draws the translation right on top of the image, in a side panel next to the page. ✅ No account, no API key, no subscription and no limits. Install it, click a picture, read it in your language. 🔒 Your pictures never leave your computer. Text recognition runs inside the extension on your own graphics card, and translation is done by Chrome's built-in on-device translator. There is no server — so there is nothing to upload, nothing to pay for, and it keeps working offline. 🎯 WHO IT IS FOR • Manga, manhwa and manhua readers — vertical Japanese text in speech bubbles, Korean webtoons, Chinese comics. • Online shoppers — Chinese, Japanese or Korean product photos, size charts and labels in online stores and marketplaces. • Social media — memes, screenshots and image posts in feeds, forums and web chats. • Travelers — photos of street signs, menus, timetables and notices. • Students and professionals — charts, slides, infographics and scanned documents in a language you don't read. It is not tied to one website: if a picture is on your screen, it can be translated. 🖱️ HOW TO USE IT • Select an area — press Alt+Shift+S (⌥⇧S on Mac) and drag a frame over any part of the page: a speech bubble, a chart, a frame of a paused video. • Capture the tab — Alt+Shift+A translates everything visible on the screen. • Right-click any image → "Translate image" — the picture is taken in full resolution, even if the page shows it small. • Paste or drop — paste a screenshot from the clipboard (Ctrl+V / ⌘V) or drag an image file into the panel. The panel shows the picture with the translation drawn over the original text, keeping its colors, angle and layout. Press and hold the picture to see the original. Below it, each text block is listed with its translation and the original; hover one to highlight it on the picture, or copy the whole translation with one click. ⭐ WHAT IT DOES WELL • Understands vertical Japanese and Chinese text, rotated and tilted signs, and pages with several columns. • Joins the lines of a paragraph before translating, so sentences are translated whole, not line by line. • Detects the language of the picture by itself — or you can choose it. • Takes right-clicked images in full resolution; very large pictures are handled too. • Works offline once Chrome has the language packs you use. 🌍 LANGUAGES Reads text in 36 languages: English, Russian, Ukrainian, Bulgarian, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Czech, Slovak, Slovenian, Croatian, Hungarian, Romanian, Lithuanian, Danish, Swedish, Norwegian, Finnish, Turkish, Indonesian, Vietnamese, Greek, Chinese (Simplified and Traditional), Japanese, Korean, Arabic, Hindi, Marathi, Thai, Tamil and Telugu. Translates into 39 languages — all of the above plus Hebrew, Bengali and Kannada. 📏 MEASURED, NOT PROMISED We test every version on 648 test images in 27 languages, made from the FLORES-200 translation benchmark, and on 70 real scanned forms from the XFUND dataset. On a MacBook Air M3 in Chrome 155: • Speed: a typical picture is recognized and translated in about 0.3 s (median). Recognition alone takes about 0.1 s on the graphics card and about 0.3 s on the processor. A whole page of a scanned form takes about 2.5 s. • Recognition: in the 14 tested languages written in Latin, Cyrillic or Greek letters (Vietnamese aside — see Limitations), and in Chinese and Japanese, 99.3% of printed characters are read correctly on average. On the real scanned forms — Chinese, Japanese, German, French, Spanish, Italian and Portuguese — 95.6% of words are found and read correctly. • Translation: averaged over the 25 tested languages it reads, translation from the picture scores 46.3 chrF++, against 49.1 for the same text with no recognition errors — reading the image costs less than 3 points. • Size: about 120 MB, recognition models included. Each translation language pack is about 80 MB, downloaded by Chrome once. ⚙️ UNDER THE HOOD • Text recognition: PaddleOCR models (PP-OCRv6 and PP-OCRv5) packed inside the extension and run by ONNX Runtime on your graphics card through WebGPU, or on the processor without one. • Translation and language detection: Chrome's built-in on-device Translator and Language Detector (Chrome 138 and newer). • No servers, no analytics, no ads. The only thing the extension downloads is the picture you choose, from the address the page already loads it from; language packs are downloaded by Chrome itself. ☁️ HOW IT COMPARES WITH CLOUD IMAGE TRANSLATORS • Where your picture goes: here it stays on your computer; cloud translators upload it to their servers. • Account, API key, payment: none here; cloud tools often require one or limit free use per day. • Offline: works here; cloud tools need the internet. • Where cloud tools are still better: handwriting, scripts we don't read yet (Hebrew, Bengali, Kannada) and Vietnamese tone marks. ⚠️ LIMITATIONS • Hebrew, Bengali and Kannada text cannot be read yet (translation into these languages works). • Vietnamese tone marks are not always recognized; complex scripts such as Arabic, Thai or Telugu occasionally lose a letter. • Handwriting is recognized only partially. • Chrome does not allow capturing its own pages (chrome://), the Chrome Web Store or DRM-protected video. 💻 SYSTEM REQUIREMENTS • Desktop Chrome 138 or newer on Windows, macOS or Linux. • A graphics card with WebGPU is recommended (most laptops of recent years); without one, recognition runs on the processor. • About 120 MB for the extension and about 80 MB for each translation language you use. 🚀 GETTING STARTED On first launch, a short tour translates a sample sign and shows where pictures come from. You can also download Chrome's language packs in advance there, so later translations start instantly. ❓ FREQUENTLY ASKED QUESTIONS How do I translate text in an image in Chrome? Right-click the image and choose "Translate image", or press Alt+Shift+S and select the part of the page with the text. The translation appears over the picture in the side panel. Is there an image translator that works offline and doesn't upload my pictures? Yes — this one. Recognition and translation run on your computer; after Chrome has downloaded the language packs, it works without the internet. Can it translate manga and webtoons? Yes. It reads vertical Japanese text in speech bubbles in the right order, as well as Korean and Chinese comics. Select a panel or a whole page. How do I translate Chinese text on product photos in an online store? Right-click the product photo and choose "Translate image". The photo is taken in full resolution, so small print on labels and size charts is readable. Can I translate a screenshot? Yes. Copy it and press Ctrl+V (⌘V on Mac) in the panel, or drag the image file into it. Is it free? Are there limits? Yes, it is free, with no limits, no account and no API key. Everything runs on your own computer, so there is nothing to meter. Why does Chrome ask me to download a language pack? Translation is done by Chrome's built-in translator. The first time a language pair is used, Chrome downloads its pack and keeps it — after that, the pair works offline. Chrome only allows this download after a click, so the panel shows a button for it. Why is the extension about 120 MB? That is the text recognition models — they live inside the extension so that nothing has to be sent to a server.