Video info

1. Video info#

Everything on the editing page is one phrase of one caption. video info, in the player’s bar, is the video itself: the fields of its video.json that YouTube’s lookup or an LLM’s answer can get wrong — a mistitled video, a channel name YouTube would not give — and that had no other way to be mended than editing the file.

save (or Ctrl+↵) writes video.json, says saved ✓ and reloads the page, since the title and the channel are shown in several places at once; nothing to change is all it says when nothing changed. ✕, Esc or a click beside the sheet close it. The video pauses while the sheet is open and plays on after, if it was playing. There is nothing to rebuild: the player reads video.json on every visit.

What it refuses: title may not be blank, and a field holding a control character — … carries U+0007 at character 3, which is not text — which a paste from a strange source can carry. The server answer, if the server is not there: the server did not answer (…) — nothing was written.

Kanbun#

A Japanese or Chinese phrase carries its words with their readings, and normally those readings run together are the phrase’s own reading — so the word strip warns when they disagree. A text read out of its written order — classical Chinese read as Japanese, kanbun — rightly disagrees everywhere. Tick read out of its written order (kanbun): it writes "reorders": true into video.json, and the readings are no longer compared with the words. Untick it and the key is taken out again. The reading editions’ book info sheet has the same checkbox for a book.

2. A video still being glossed#

Nothing marks a video as unfinished. A video started empty is an ordinary video whose phrases have nothing written in them yet, and a phrase nobody has glossed is legal in every video, for as long as it stays so. There is no flag to set and none to take off when the last one is written — a "draft": true an older Parseh left in a video.json is never read, and may stay or go.

Where it shows:

• on the video’s card, phrases are counted rather than captions while any is blank — 12 of 40 chunks glossed — since a video started empty has a phrase under every caption and nothing written in them (The video index and channels);
• in the player every blank phrase of the language is a phrase like any other — hoverable, with a cloud that says nothing glossed yet and a ✎ — because it is exactly what you opened the page to fill in (The video player).

It is glossed a phrase at a time in the ✎ form (Editing a phrase), where a meaning may be saved before its transliteration, or a run at a time by an LLM (Glossing captions with an LLM), which fills only the phrases nobody has glossed.

What the checker makes of it. A phrase with nothing written in it — no transliteration, no vocabulary, no meaning, no kana — is counted, once for the whole video (note: 12 of 40 chunks have no gloss yet), and asked for nothing else. A phrase half written — the meaning typed, the transliteration still to come — is an error to the checker, which is how the half-done work is found; the ✎ form saves it on the way, and a bundle that carries one is still taken back in, with a note saying how many.