VIDOUMOビドウモウ

• On and kun readings. Every kanji on the card now lists its readings, and the one the word uses is picked out in green. In 食堂 it is しょく and どう, so you can see the しょく you will meet again in 食事. Where a word uses a reading the lists do not cover, as 日 does in 日本語, nothing is picked out.
• The radical, named. The stroke count is gone. The radical now carries its Japanese name in romaji, tsuchi for 土, which cannot be mistaken for one of the kanji’s own readings. A kanji that is its own radical, like 食, does not list itself.
• A wider card. It is a little wider so the readings fit, and its scrollbar keeps clear of the rounded corners.

Earlier

• v1.5.0 · 16 Oct 2026 · Every kanji,. Kanji on the card. The dictionary card shows each kanji in the word you clicked: its radical, what it means on its own and how many strokes it takes. A single kanji gets the width of the card to itself, and a long word trims each meaning to fit. All 6,355, built in. Every kanji in the JIS X 0208 set is in the package, so the table appears with the entry and nothing extra leaves your browser. The data is KANJIDIC2 and KRADFILE, credited in LICENSES.md. The reading as it was said. The card now opens with the caption’s own reading in hiragana, with romaji beside it. A conjugated word says what it is, as in “行かなく is the negative form of 行く”. A name with no dictionary entry, like 深谷, still gets its kanji and can be saved. Better matches. The entry follows the reading in the caption, so 日 read ひ and 何 read なん give the right word. A kana word goes to the one usually written in kana. 行った is 行く unless the sentence is formal, and a lone kana or a stray fragment is no longer clickable.
• v1.4.0 · 13 Oct 2026 · Now on. Netflix works. Pick a Japanese subtitle in the Netflix player and it becomes clickable words with furigana, English underneath and the dictionary card, as on YouTube. Saved words keep the title and link back to the moment. English from Netflix first. If the title has English subtitles, those go under the Japanese. If not, a line goes to Google, which is the one thing that leaves your browser. The level shows on Netflix too. The JLPT pill sits in front of the title in the player controls. Speed is judged over the whole running time, on a scale I have guessed and not fitted, so it moves the result very little. A new permission. Running on netflix.com is a new site permission, so Chrome will ask existing users to accept it before the extension switches back on.
• v1.3.0 · 10 Oct 2026 · Grammar now. Grammar counts. The level estimate now looks for N3, N2 and N1 grammar patterns as well as words and kanji. Vocabulary alone cannot tell a beginner conversation from an intermediate one, because they use the same everyday words. Hover the pill for the new Grammar row. Easy videos read as easy. The estimate used to squeeze nearly everything into N3 and N4. A slow beginner podcast now reads N5 and a university lecture N1 or N2. It says when it does not know. A video with too little speech now reads “JLPT unrated” and says why, instead of guessing from a handful of words. Every rule is written down. How the estimate works, where each number comes from and how far to trust it is at /vidoumo/levels.
• v1.2.4 · 8 Oct 2026 · How hard. A JLPT level in front of the title. A small green pill, for example “JLPT N3–N4”, sits before the video’s title. It is an estimate from three things in the Japanese captions: how hard the words are, how hard the kanji are, and how fast people speak. A “~” means the captions are automatic, so it is less certain. Hover it for the breakdown. The words and kanji split by level, and where the speech sits on a slow-to-fast line. It is worked out in your browser, so nothing extra leaves it. It is a guide, not a rating. The JLPT publishes no word or kanji lists, so the levels come from Jonathan Waller’s unofficial ones, and grammar is not counted. Automatic captions hold steadier. A line with no timing of its own now stays up for as long as its words suggest instead of flashing and vanishing, and a stray one-word tail of English no longer appears on its own.
• v1.2.2 · 7 Oct 2026 · Tidied. Nothing changes on screen. The unused end-of-video quiz code is gone from the package, so it contains only what the extension actually does. Licences are in the package. A new LICENSES.md carries the notices for kuromoji, its IPADIC dictionary and the other code bundled with it.
• v1.2.1 · 6 Oct 2026 · English from. The uploader’s English comes first. If the video has English subtitles, they now sit under the Japanese, written by a person and timed to the video, instead of a machine translation of it. Then YouTube’s own translation. With no English track, Vidoumo asks YouTube for its translation of the Japanese, rebuilds it into whole sentences and shows the sentence being spoken. Google is the last resort. Only if neither is available does a line go to Google, line by line as before. When the English comes from YouTube, nothing is sent to Google. English stays with its line. It arrives with the Japanese rather than a beat after, and stays up until the line ends. Hover the English to see where it came from.
• v1.1.1 · 6 Oct 2026 · Captions that. Captions no longer type out. Vidoumo now reads the whole caption track YouTube downloads, not just what is on screen at that moment, and shows each phrase complete at its time. The box stops resizing and flickering while a line arrives. Furigana and translation are ready first. The next few lines are translated and split into words ahead of time, so they appear finished instead of filling in a moment later. Automatic captions are rebuilt into phrases. YouTube delivers them in short rows, often mid-sentence. They are joined back into sentences and paced to a normal speaking rate, so a phrase does not run ahead of the speaker, and sound notes like [音楽] stand on their own. Quiet after an update. Reloading the extension with a YouTube tab open used to throw “Extension context invalidated” errors and leave YouTube’s own captions hidden. It now steps aside cleanly. Live streams and anything it has not captured still use the earlier method of reading the screen, so nothing stops working.
• v1.0.9 · 6 Oct 2026 · First public version. Furigana, translation, dictionary cards, saved words with a link back to the moment, and CSV export.