Videos

The Videos door is the reading editions’ method applied to a video. The video plays at the top of the page and its whole transcript runs underneath it, cut into small phrases. Point at a phrase and a cloud opens with what a reading edition would print beside it: the transliteration (and, for Japanese, the kana above it), the vocabulary, the meaning. The line being spoken is lit as the video plays, and a click on any line plays it again from its beginning — as many times as you like, which is the whole point.

It works for every language Parseh teaches — Persian and Arabic right to left, Italian, French, German, Spanish, Turkish and English in Latin letters, Hindi in Devanagari, Japanese and Chinese with the reading set over each word — and the meanings are written in whatever language the video says they are: English unless it says otherwise, Italian for an Italian learning English, Persian for a Persian learning Japanese.

A video is either on YouTube, and streams from there, or a film on this machine, which needs no internet at all. Either way the transcript, the glosses and everything you write are files in Parseh’s own folders, read afresh on every visit: nothing is built, and nothing has to be restarted.

1. Where it is#

Open the hub and press Videos, or go straight to https://localhost:7654/youtube/. The hub’s door counts the videos and the channels, and the language chips at the top of the hub narrow the count to one language. In Mobile mode the door opens the same pages: the videos have no mobile version of their own yet.

2. The short version#

Find a video: pick a channel on the videos page, then a video.
Watch and read: hover a phrase for its gloss; click a line to hear it again; turn on Pause while a gloss is open (in the ⚙ panel) to have the video wait while you read, and press L on a phrase to flag it for later (Review later).
Add a video: the ＋ Add a video card asks where the video is and who writes the glosses — an LLM whose answer the page checks, or you, in the player.
Write it: the ✎ in any gloss cloud opens the phrase’s fields; gloss with an LLM has a chatbot gloss a run of captions; the timings moves where a caption starts; a + between two lines holds a note of your own.
Take it away: the ⤓ in the player’s bar downloads the video as one zip that another Parseh installs whole.

3. The pages of this section#

The video index and channels — the videos page, the channels, the cards and what their tags say, taking a video off the shelf.
The video player — every control in the bar and the ⚙ panel, reading a transcript, beside the text, the grip, review later, phones.
Reading help — the dictionary under a phrase nobody has glossed, the definitions, the readings of Japanese and Chinese, and Decompose Kanji.
Adding a video — the two questions, the prompt for any LLM, checking its answer, and starting a video empty.
Mending the transcript first — the small subtitle editor on the add page: timings to a tenth of a second, ✨ tidy up, and the same job done by a model.
A film on this machine — a video that is a file, a film or a sound alone: sending it instead of naming it, subtitles, the hardlink, playing offline, downloading it with its film.
Editing a phrase in the player — the four colours, the ✎ form, the sources beside it, and every refusal.
Cutting and joining phrases — moving where a phrase ends.
Glossing captions with an LLM — a run of captions glossed by any chatbot, only where nobody has glossed, and deleting a gloss to have it written again.
Notes between the captions — a studio document in the seam between two lines.
The timings — moving where a caption starts, over a picture of the sound.
Video info — the title, the channel and the level, and a video whose phrases are still being glossed.
Taking videos away and bringing them back — the ⤓ bundle and its size, bringing one back, the backup of the whole shelf.
A video’s files — video.json, annotations.json, transcript.txt and the rest, field by field.

Cards — Anki notes, exercises for a deck, studio markdown — are made from the player too (Alt-click a word, or + card in a cloud); the section Cards and Anki explains the sheet that opens.

The index and channelsThe videos page, the channels, what the tags on a card say, and taking a video off the shelf.
The playerReading a transcript while the video plays, every control in the bar and the ⚙ panel, beside the text, the grip, review later, and phones.
Reading helpThe dictionary under a phrase nobody glossed, the definitions, the readings of Japanese and Chinese, and Decompose Kanji.
Adding a videoThe add page's two questions, a transcript pasted or made on this computer by speech to text, the prompt for any LLM and the check of its answer, and a video started empty to gloss by hand.
Mending the transcriptThe add page's small subtitle editor — recorded word starts where Parseh made the transcript, timings to a tenth of a second, cutting and joining captions, ✨ tidy up in two modes, and the same job done by an LLM.
A film on this machineA video that is a file rather than a YouTube address, a film or a sound alone — named by its path or sent, its subtitles, the hardlink, playing offline, and taking it away with it.
Editing a phraseThe four colours, the ✎ form and its fields, the note, deleting a gloss, the words of Japanese and Chinese, the sources beside the fields, and every refusal.
Cutting and joiningMoving where a phrase ends — cut in two, join next, join previous — what is proposed for each half, and what is refused.
Glossing captions with an LLMThe player's gloss with an LLM panel — picking a run of captions, the two checkboxes, the prompt, the answer pasted back, the report — and why a gloss already written is never touched.
Notes between captionsA studio document written into the seam between two captions — writing one, reading one, where it says it sits, and where it lives.
The timingsThe one door that moves where a caption starts after a video is added — the sheet, its keys, estimating the rest — or only the next few seconds — by the text or by the sound, the picture of the sound, and why it moves nothing else.
Video infoThe video info sheet — title, channel, level, blurb, the kanbun checkbox — and a video whose phrases are still being glossed.
Downloads and backupsA video's ⤓ bundle and its size, bringing one back with its refusals, the backup of the whole shelf, and the trash.
A video's filesWhat a video is on disk — video.json, transcript.txt, annotations.json, parts, notes, the waveform and the film — field by field, and the tools behind the pages.