Translates and explains Japanese, Chinese and Korean from anything on your screen. Free, and without leaving your browser.
Add to Chrome · FreeDesktop Chrome and Edge
HOW IT WORKS
No copy and paste, and no matter how the page is built.
STEP 1
Open KomaLens from its icon or the side panel and draw a box over the speech bubble, subtitle or sign.
STEP 2
It captures what is on screen and recognizes the text with OCR. In local mode, the image never leaves your browser.
STEP 3
Get the translation plus every word with its reading, meaning and type.
WHY IT’S DIFFERENT
Pop-up dictionaries and page translators need real text inside the HTML. When the text is inside an image, a canvas or a video, they can’t find it. KomaLens captures the screen: if you can see it, it can read it.
And no hassle: it’s free, installs from the Chrome Web Store in one click, reads Japanese, Chinese and Korean, and exports to Anki. Other options charge you, need extra software or stick to a single language.
Works with
| Tools that read the code | KomaLens | |
|---|---|---|
| Regular text on a website | Yes | Yes |
| Scanned pages and images | No | Yes |
| Readers that draw on a canvas | No | Yes |
| Subtitles and signs inside a video | No | Yes |
| Sites that block text selection | No | Yes |
MODES
Three modes for three needs. Switch anytime from the mode chip in the panel or from Settings.
Free · Private · No API key
OCR with PP-OCRv6 and translation with the browser’s built-in Translator API. Everything happens on your device.
Low cost · Fast
OCR runs on your device and only the recognized text goes to your AI provider. Faster and cheaper than sending the image.
Best quality
The cropped image goes to a vision model from your provider. For tough pages, hand-drawn lettering or heavily stylized text.
Hybrid and Image to API work with any provider that follows the OpenAI standard, or with a model running on your own computer.
You only pay for what you use, directly to your provider. KomaLens charges nothing.
FEATURES
It doesn’t just translate: it shows you how each sentence is put together and helps you review it later.
The sentence is underlined in the color of each word type: nouns, verbs, adjectives, expressions, proper names and particles. Every word comes with its reading and meaning, and you can hide particles to focus on what matters. In local mode, segmentation and readings are computed in your browser.
Read the guide : Word by wordSave words with the star, filter by language and export everything to a CSV with the columns Char, Read, Meaning and Type, ready to import into Anki.
Read the guide : My list and AnkiJapanese, Chinese and Korean work in Local, Hybrid and Image to API, each with its own reading system: kana, pinyin and romanization.
Read the guide : Three languages, in every modePaste the base URL, your API key and the model ID. Any OpenAI-compatible API works, even a model running on your computer with Ollama or LM Studio. We don’t push any model on you: pick the one you want.
Read the guide : Your provider, your modelIf something fails, KomaLens tells you what happened, why, and how to fix it, with a button that does it. If the browser won’t let it read a tab, “Allow on this site” asks for permission for that site only.
Read the guide : Errors that explain themselvesSTRENGTHS
What you get from the very first scan.
No paid plan and no scan limit in local mode.
In local mode, the image never leaves your browser.
Install it and use it. No sign-up, no login.
Images, canvas, video and sites that block text selection.
Use the account you already have with any OpenAI-compatible API, or a local model.
Your data goes straight to your provider, never to us.
Words with readings and types, plus Anki export.
Models and dictionaries ship with the extension. After installing, it downloads nothing from third parties, except the browser’s own language packs.
PRIVACY
KomaLens has no servers. What is never sent can never leak.
Read the privacy policyCOMING SOON
See the translation just by moving the cursor over areas with text.
Translate the whole page on your device and replace the text on screen. It’s the goal for version 3.
KomaLens is made by a developer learning Japanese and Chinese, and it grows with feedback from the people who use it. Want another language or a new feature? Let me know.
FAQ
Yes. KomaLens is free, with no paid plan and no scan limit in local mode. If you use Hybrid or Image to API, you pay your AI provider directly for what you use, at their rates. KomaLens charges nothing and takes no cut.
Not for local mode, which is the default. You only need one for Hybrid and Image to API, and you create it in your provider’s dashboard. With a local model in Ollama you don’t need one either.
Local does everything in your browser: free and private, with translation and words. Hybrid reads the text on your device and sends only that text to your provider, which adds notes and a literal translation. Image to API sends the cropped image to a vision model: the highest-quality option for tough pages.
Not in local mode: the capture is processed in your browser. In Hybrid only the recognized text is sent, and in Image to API the cropped image is sent, always straight to your provider. Nothing goes through KomaLens servers, because there are none.
It reads Japanese, Chinese and Korean in every mode, and for now translates into Spanish or English. Missing a language? Let me know: suggestions are welcome.
Yes, through that provider’s API: create an API key in their dashboard and paste it into Settings. Keep in mind that many providers bill API usage separately from their chat subscription; for example, OpenAI bills ChatGPT subscriptions and API usage separately. Check how your provider bills.
Whichever you like: we don’t recommend any. In Hybrid, any text model works. In Image to API the model must accept images; your provider lists this on the model’s page.
Yes. With Ollama or LM Studio, KomaLens connects to your own computer and nothing leaves it. For Image to API you need a vision model.
On almost any: web pages, images, canvas readers, video and sites that block text selection. The browser doesn’t let any extension read its internal pages (like chrome:// or edge://) or the extension store, so those can’t be scanned.
Because KomaLens only asks for what it needs, when it needs it. If the browser won’t let it read the tab, the “Allow on this site” button requests access to that site only. When you set up a provider, it asks to connect to that domain only, or to localhost if you use a local model. If you decline, everything else keeps working.
Save words with the star, open My list and click “Export to Anki”. You get a CSV with the columns Char, Read, Meaning and Type, which you import into Anki from File → Import.
Desktop Chrome and Microsoft Edge. Local mode needs the browser’s built-in Translator API: it is tested from Chrome 156.0.8078.47 and should work from the first stable Chrome version that includes it. Other Chromium-based browsers might work if they ship that API, but they are untested. Hybrid and Image to API don’t depend on it. It doesn’t work on mobile.
Still have questions? Write to poullorca@proton.me or to @PoulLorca on X.
Installs in a minute. No account, no credit card.
Add to Chrome · Free