get_timedtext.js

const fetch = require('node-fetch')
const { JSDOM } = require('jsdom')
const nodeEval = require('node-eval')
const xmlParser = require('fast-xml-parser')

const DEFAULT_LANG = 'ja'

function extractTracksFromHTML(body) {
    const dom = new JSDOM(body)

    const context = {
        window: {},
    }
    Array.from(dom.window.document.querySelectorAll('body script'))
        .map(node => node.textContent)
        .filter(text => text.trim().startsWith('window["'))
        .forEach(text => {
            // console.log(text)
            nodeEval(text, 'watch.js', context)
        })

    const captions = context.window.ytInitialPlayerResponse.captions
    const tracks = captions.playerCaptionsTracklistRenderer.captionTracks

    return tracks
}

function selectTrack(tracks, lang) {
    const candidates = {}
    tracks.forEach(track => {
        if (track.languageCode === lang) {
            const key = ( track.kind === 'asr' )
                  ? 'generated'
                  : 'manual'
            candidates[key] = track
        }
    })
    const track = candidates.manual || candidates.generated
    return track
}

function extractTextsFromXML(body) {
    const data = xmlParser.parse(body)
    const texts = data.transcript.text
    return texts
}

function main() {
    const videoId = process.argv[2]
    const lang = process.argv[3] || DEFAULT_LANG

    fetch(`https://www.youtube.com/watch?v=${videoId}`)
        .then(res => res.text())
        .then(body => {
            const tracks = extractTracksFromHTML(body)
            const track = selectTrack(tracks, lang)
            fetch(track.baseUrl)
                .then(res => res.text())
                .then(body => {
                    const texts = extractTextsFromXML(body)
                    texts.forEach(text => console.log(text))
                })
        })
}

if (require.main === module) {
  main()
}