Japanese 字幕 files with the timings solved.

Japanese is written without spaces between words, which removes the natural break points a Latin-script caption relies on. A renderer that wraps wherever the line runs out will happily split a word, or separate a particle from the phrase it marks. Japanese captioning practice breaks at phrase boundaries instead — after a particle such as は, が, を or に, or between clauses — so each line is a unit a reader can take in at a glance.

Character density changes the line budget completely. Broadcast practice in Japan has long worked to roughly thirteen full-width characters a line, two lines at a time, because each character carries far more information than a Latin letter and takes about twice the width. Porting an English cue sheet's forty-two-character lines into Japanese produces lines nobody can read in the time available.

• Full-width punctuation 、 and 。 belongs in the cue text, and a line does not begin with a closing bracket or a small kana.
• Three scripts run together in one sentence — kanji, hiragana and katakana — and switching between them is normal, not an error to fix.
• Vertical writing is common in print but SRT and VTT are horizontal formats, so cues are laid out horizontally.
• Katakana marks loanwords and emphasis, which is information a viewer reads; converting it to kanji would change the tone.

Nothing in the downloaded SRT or VTT labels the speaker — both are cue text and timecodes. The transcript view separates voices on screen, and the JSON export from the same run keeps that as data.