Open a Japanese webpage
Use it on ordinary news, forum, academic, or sports pages. The extension adds help directly to the current page without replacing the entire page body or its links.
Chrome Extension v2.6.0
Read Japanese websites,
starting with every kanji.
Open news, forums, papers, or sports coverage and add readings above kanji with one click. You can separately reveal the origins of katakana loanwords and show supported foreign personal and place names in their original scripts, then translate a selected word, sentence, or paragraph through your own API.
- Readings processed locally
- Loanword origins matched locally
- Only text you select is translated
- No ads or tracking scripts
テクノロジー
日本の 研究チーム、 新技術を 発表
最新の 衛星画像から、 台風の 進路を 分析しました。
Selection translationSentence
We analyzed the typhoon's projected path using the latest satellite images.
Readings are enabled on this page
Local readingsReading analysis stays on your device
LocalJapanese reading analysis
4 languagesChinese, English, Japanese, and Korean UI
DynamicHandles content loaded later
Chrome 120+Manifest V3
Overview
Less time breaking your flow to look things up
This is not a replacement reader. It is a lightweight Japanese aid that you can add to or remove from the current page at any time.
Turn on readings
The extension analyzes text locally in your browser and displays readings above kanji with standard ruby markup.
Translate only when needed
Select a word, sentence, or paragraph and view its translation on the same page without repeatedly switching tabs.
Features
Designed for reading real webpages
From ordinary kanji and loanword origins to news dates, names, and long-sentence translation, every feature is built to preserve the original page structure.
01 · Smart readings
See readings above kanji at a glance
Kuromoji performs morphological analysis locally in the browser. Content loaded later is also processed, and clicking again removes the readings added by the extension.
- Skips code, forms, and existing ruby markup
- Conservative disambiguation for dates, loanwords, and common contexts
- Adds readings only to kanji and keeps okurigana intact
- Shrinks long body-text readings while keeping header and navigation widths stable
- Avoids rearranging element hierarchy to preserve styles and interactions
After annotation
東京で 新しい 技術を 発表
麻辣湯読みます
02 · Loanword origins
See original foreign words above katakana
The popup includes a separate “Show loanword origins” button. When enabled, the extension uses 3,353 bundled local etymology records to match high-confidence loanwords and display original forms for well-known foreign leaders, public figures, and places. It makes no API call and does not upload page text.
- English entries show only the original word, without a country marker
- Other languages use Japanese kanji country abbreviations such as 仏, 独, and 西
- Long annotations collapse to the katakana base width and expand in a width-limited popover
- Full proper names can match directly; shortened names require role, country, or person context
- Homographs, wasei-eigo, and unknown words are conservatively skipped
Etymology data is derived from JMdict under the EDRDG General Dictionary Licence Statement and CC BY-SA 4.0.
With loanword origins enabled
computerコンピューター
仏coup d'Étatクーデター
独Arbeitアルバイト
西paellaパエリア
Long original forms appear as a collapsed summary. Click or press Enter or Space to see the complete form in a popover without expanding line spacing.
Translate words, sentences, or paragraphs
Select the part you need help with and open a translation card on the original page, with copy and retry controls. When the API returns usage data, the card shows the tokens used by that request.
新しい研究成果が発表されました。
Selection translationSentence
New research findings have been announced.
English · 12 tokensUse verified readings for public figures without guessing ordinary names
Registered full-name readings for Japanese public figures take priority over incorrect guesses from webpages or the tokenizer. Foreign names, news dates, people counts, and academic terms use constrained matching; uncertain text remains unchanged.
24日（月）にじゅうよっか・げつ
大谷翔平おおたに・しょうへい
羽生結弦はにゅう・ゆづる
Translation dashboard
Manage your API, languages, history, and usage in one place
Connect your own OpenAI-compatible service. The interface language and translation target remain independent, so every setting stays clear and controllable.
API
Custom connectionAPI key, Base URL, and model
A
Target language8 common translation targets
H
Translation historySearch, delete, and clear locally
Σ
Token recordsInput, output, and daily totals
JAPANESE READING ASSISTANT
API settings
API Key••••••••••••
Base URLhttps://api.example.com/v1
Model nameyour-model-name
Privacy
Reading analysis stays local; you decide what gets translated
Japanese readings and loanword-origin matching always happen locally in your browser. Of the page content, only text you actively select is sent to your configured translation service. Testing the connection sends the fixed test text “こんにちは”.
- 01
API Key is stored on your deviceStored in a trusted extension context and never shown in full
- 02
API permission is requested as neededNo permanent permission to access every website API
- 03
You control your historyTranslation history and token records can be cleared at any time
Installation
Start reading Japanese webpages in minutes
This version is installed through Chrome's “Load unpacked” flow and requires Chrome 120 or later.
Official unpacked installation packageJapanese Kanji Reading Assistantv2.6.0 · ZIP · 14.9 MiB · Chrome 120+
SHA-256D9F1D1D8F6A7B98AEC19DB91B562FA7A5B1589681CFB07C670C66C17AD755CD4
- 01
Download and extract
Download the ZIP above and extract it. Confirm that the resulting folder contains
manifest.json. - 02
Open the extensions page
Enter
chrome://extensionsin Chrome's address bar and turn on “Developer mode”. - 03
Load the unpacked extension
Click “Load unpacked” and select the extension folder you just extracted.
- 04
Open a page and enable it
Open a Japanese webpage, click the “漢” icon in the toolbar, then enable readings, loanword origins, or selection translation as needed.
You do not need an API Key to use readings. For selection translation, enter the API key, Base URL, and model for your compatible service. When updating, click “Reload” on the extensions page and refresh the webpage; do not remove the extension first.
FAQ
What you may want to know before starting
The extension is local-first and connects only when needed, so readings and translation can be used independently.
Do readings require an API?
No. Kanji readings are analyzed locally in your browser. Only selection translation requires your own compatible API.
Does the extension upload the entire webpage?
No. Reading annotations do not use the network. Of the page content, only text you actively select is sent to your configured translation service. Clicking “Test connection” sends the fixed text “こんにちは”. The provider processes data under its own privacy policy.
How do I remove readings from the current page?
Open the extension popup again and click “Remove readings”. The extension removes its generated ruby markup and restores the original text, while selection translation remains available. A small number of highly dynamic pages may require a refresh.
Why do some katakana words have no original form?
Katakana can also represent wasei-eigo, onomatopoeia, or proper names. The extension annotates only verified words with clear origins, using a fixed high-confidence JMdict subset dated 2026-08-17, a manually reviewed common-word layer, and selectively checked National Institute for Japanese Language and Linguistics vocabulary lists. Cross-entry homographs, unknown words, and shortened names without safe context remain unchanged. English shows only the original word; other languages and foreign proper names use Japanese country abbreviations such as 仏, 独, and 西.
Which pages cannot be processed yet?
Extensions cannot be injected into Chrome system pages. Closed shadow DOM, cross-origin iframes, rare characters, and unknown proper names may remain unchanged. Code blocks, forms, existing ruby markup, and editable areas are deliberately skipped.
What must a translation API support?
The API must support POST /chat/completions. Remote services must use HTTPS; only localhost and 127.0.0.1 may use HTTP. Each request is limited to 12,000 characters and depends on your own key, network, quota, and model access. Local token records are estimates and are not the provider's bill.
Reading Japanese webpages can feel more natural.