Sound / 音
Kanjium accent DB (CC BY-SA) → drawn + hummed locally
A learner connecting ideas to sounds needs to see things. This page is a working proof: every widget below is fed live by a community resource — pitch accent shapes from Kanjium, stroke order you can trace, real anime lines with screenshots, pictures resolved from Wikidata. One tap on a lyric word lights all of it up.
This is the opening line of 故郷 (1914) — public domain, so we can print it freely. Tap any underlined word and the dock below fills in: how it sounds, how it's written, what it means, what it looks like, and where it shows up in real shows.
The whole line as one continuous high/low contour — mora by mora, word by word. Press play: the dot rides the ribbon, each mora lights as it passes, and a two-tone hum (WebAudio, no recordings needed) lets you hear the shape before you ever sing it. This is the lyric-page overlay I'd ship.
Word-level accents from Kanjium; the join between words is illustrative (real phrase contours need accent-phrase rules — that's what UniDic's aType + pyopenjtalk/marine give you server-side).
Same data, different moments. A lyric student doesn't want a dictionary — they want the right sliver of it at the right beat of the song.
The contour above, rendered as a thin ribbon under the active lyric line. Accent shape becomes part of the karaoke bounce — you absorb it without studying.
What you tried above: one tap, five senses. Sound, shape, meaning, picture, real usage. Nothing leaves the page; the song keeps playing.
The line as frames: one picture per content word, in order. Watch the story of the sentence before you hear it — rabbit, mountain, fish, river.
Japanese rhythm is mora-timed — exactly what trips singers. Dots pulse one per mora at song tempo; high dots sit higher. Tap to feel 兎追いし as six even beats.
Color each lyric word by corpus frequency. Learners triage instantly: the warm words pay rent everywhere; the cool ones are this song's flavor.
Between plays, deal three anime/drama lines containing a word you just sang — screenshot, audio, source. The word stops belonging to the song and starts belonging to the language.
Pre-song ritual: trace the line's hardest kanji with your finger (quiz mode grades each stroke). Motor memory is a memory channel most apps never touch.
Some things you embed, some you respectfully link: Forvo for human voices, YouGlish for the word in YouTube speech, OJAD for conjugated accent, Massif for 30M novel sentences. Links cost nothing and never rot your legal standing.
Everything used or referenced on this page. Badge key: FREE API call it live · CC DATA download, attribute, ship · GRAY community-tolerated, link don't re-host · PAID commercial terms.