@lincyaw/addon-lecture 0.0.0-stage → 1.0.0-rc.10

package/src/domain/editing.ts

@@ -0,0 +1,112 @@
+import type { TrackEvent, TranscriptSegment } from '@lincyaw/kernel'
+import type { Clip } from './script-track.ts'
+import { anchorOf } from './timing.ts'
+import type { LectureTrack, MediaSegment, Utterance, Voice } from './track.ts'
+import { charOffset, locate, realign, reanchor, trackWords } from './words.ts'
+
+export interface SegmentTranscript {
+  segment: MediaSegment
+  segments: TranscriptSegment[]
+}
+
+function stepAt(track: LectureTrack, section: string, t: number): number {
+  let step = 0
+  for (const [at, type, data] of track.events) {
+    if (at > t) break
+    if (type === 'go' && data.section === section && typeof data.step === 'number') step = data.step
+  }
+  return step
+}
+
+function anchorAtTime(utterance: Utterance, t: number): number {
+  if (utterance.words?.length) {
+    const index = utterance.words.findIndex(([start, d]) => start + d > t)
+    return index === -1 ? utterance.words.length : index
+  }
+  const length = [...utterance.text].length
+  return Math.round(length * Math.max(0, Math.min(1, (t - utterance.t) / Math.max(1, utterance.d))))
+}
+
+export function bindEvents(track: LectureTrack): { track: LectureTrack, bound: number } {
+  const ids = new Set(track.utterances.map(u => u.id))
+  let bound = 0
+  const events = track.events.map(([t, type, data]): TrackEvent => {
+    const at = anchorOf(data)
+    if (at && ids.has(at[0])) return [t, type, data]
+    const { at: _stale, ...rest } = data
+    const utterance = track.utterances.find(u => u.t <= t && t < u.t + u.d)
+    if (!utterance) return [t, type, rest]
+    bound++
+    return [t, type, { ...rest, at: [utterance.id, anchorAtTime(utterance, t)] }]
+  })
+  return { track: { ...track, events }, bound }
+}
+
+export function transcribedTrack(track: LectureTrack, transcripts: readonly SegmentTranscript[]): { track: LectureTrack, bound: number } {
+  const said = transcripts
+    .flatMap(({ segment, segments }) => segments.filter(s => s.text.trim()).map(s => ({ segment, s })))
+    .sort((a, b) => (a.segment.start + a.s.start) - (b.segment.start + b.s.start))
+  const utterances: Utterance[] = said.map(({ segment, s }, i) => {
+    const t = Math.round((segment.start + s.start) * 1000)
+    const words = trackWords(s.words.map(w => ({ text: w.text, start: w.start - s.start, end: w.end - s.start })), t)
+    return {
+      id: `u${i + 1}`,
+      t,
+      d: Math.max(1, Math.round((s.end - s.start) * 1000)),
+      text: s.text.trim(),
+      ...words.length ? { words } : {},
+      alignment: words.length ? 'words' : 'none',
+      section: segment.section,
+      step: stepAt(track, segment.section, t),
+    }
+  })
+  return bindEvents({ ...track, utterances })
+}
+
+function moveAnchors(events: readonly TrackEvent[], id: string, anchor: (offset: number) => number): TrackEvent[] {
+  return events.map(([t, type, data]): TrackEvent => {
+    const at = anchorOf(data)
+    return at && at[0] === id ? [t, type, { ...data, at: [id, anchor(at[1])] }] : [t, type, data]
+  })
+}
+
+export function editUtterance(track: LectureTrack, id: string, text: string): LectureTrack | undefined {
+  const index = track.utterances.findIndex(u => u.id === id)
+  if (index === -1) return undefined
+  const { clip: _clip, ...before } = track.utterances[index]!
+  const { utterance, anchor } = realign(before, text.trim())
+  const utterances = track.utterances.map((u, i) => (i === index ? utterance : u))
+  return { ...track, utterances, events: moveAnchors(track.events, id, anchor) }
+}
+
+export function resynthesizedTrack(track: LectureTrack, clips: ReadonlyMap<string, Clip>, voice: Voice): LectureTrack {
+  let events = track.events
+  let end = 0
+  const utterances = [...track.utterances].sort((a, b) => a.t - b.t).map((said): Utterance => {
+    const clip = clips.get(said.id)
+    const t = Math.max(said.t, end)
+    const shift = t - said.t
+    const u = shift ? { ...said, t, ...said.words ? { words: said.words.map(([at, d, w]): [number, number, string] => [at + shift, d, w]) } : {} } : said
+    end = u.t + (clip ? Math.round(clip.duration * 1000) : u.d)
+    if (!clip) return u
+    const words = clip.words?.length ? trackWords(clip.words, u.t) : undefined
+    const before = u.words ?? []
+    let anchor: (offset: number) => number
+    if (words && before.length) {
+      anchor = reanchor(u.text, before, words)
+    } else if (words) {
+      const places = locate(words.map(w => w[2]), u.text)
+      anchor = (offset) => {
+        const index = places.findIndex(p => p !== undefined && charOffset(u.text, p.start) >= offset)
+        return index === -1 ? words.length : index
+      }
+    } else {
+      const places = locate(before.map(w => w[2]), u.text)
+      anchor = offset => (places[offset] ? charOffset(u.text, places[offset]!.start) : [...u.text].length)
+    }
+    events = moveAnchors(events, u.id, anchor)
+    const { words: _words, ...rest } = u
+    return { ...rest, d: Math.round(clip.duration * 1000), clip: clip.file, alignment: words?.length ? 'words' : 'proportional', ...words ? { words } : {} }
+  })
+  return { ...track, voice, utterances, events }
+}

package/src/domain/format.ts

@@ -0,0 +1,13 @@
+export function clockText(ms: number): string {
+  const total = Math.max(0, Math.floor(ms / 1000))
+  const s = String(total % 60).padStart(2, '0')
+  const m = Math.floor(total / 60) % 60
+  const h = Math.floor(total / 3600)
+  return h > 0 ? `${h}:${String(m).padStart(2, '0')}:${s}` : `${m}:${s}`
+}
+
+export function fileStem(section: string): string {
+  // eslint-disable-next-line no-control-regex
+  const stem = section.replace(/[/\\?%*:|"<>#\u0000-\u001F\s]/g, '_').replace(/^\.+/, '_')
+  return stem.slice(0, 80) || 'section'
+}

package/src/domain/index.ts

@@ -0,0 +1,50 @@
+export { dropChangedSections, trimToDeck } from './audience.ts'
+export { configuredLectureVoice, lectureGap, lectureVoice, lexiconOf, narrationVoice, prepareNarrationVoices } from './config.ts'
+export type { LectureVoice } from './config.ts'
+export { blockDrift, blockKey, blockOf, eventDrift, irHash, playedBlocks, referencedBlocks, sectionHashes, trackBlock, unchanged } from './drift.ts'
+export type { BlockDrift, EventDrift, Grade } from './drift.ts'
+export { bindEvents, editUtterance, resynthesizedTrack, transcribedTrack } from './editing.ts'
+export type { SegmentTranscript } from './editing.ts'
+export { clockText, fileStem } from './format.ts'
+export { LECTURE_MESSAGES, LECTURE_TOOL_ERRORS } from './messages.ts'
+export type { LectureMessages } from './messages.ts'
+export { CUE_COMMANDS, cueProblem, lectureRole, parseCue, spokenText } from './narration.ts'
+export type { CueCommand, CueMark, SpokenText } from './narration.ts'
+export { lecturePlan, plannedAnimations, plannedUtterances, sectionHints } from './plan.ts'
+export type { LecturePlan, PlannedAnimation, PlannedCue, PlannedParagraph, PlannedPiece, PlannedSection, PlanProblem } from './plan.ts'
+export { recordedTrack, sectionSpan, simplifyPath, spliceSection } from './recording.ts'
+export type { PathPoint, Recording, TrackHead } from './recording.ts'
+export { scriptTrack } from './script-track.ts'
+export type { Clip, CutAnimation, RecordingClip, ScriptHead } from './script-track.ts'
+export { anchorOf, lastIndexAt, stepStops, textPieces, timedEvents, trackChapters, utteranceAt, wordTime } from './timing.ts'
+export { captionGroups, captionGroupAt, trackCaptionGroups } from './captions.ts'
+export type { CaptionGroup } from './captions.ts'
+export type { StepStop, TextPiece, TimedEvent, TrackChapter } from './timing.ts'
+export {
+  audioSegments,
+  CLIP_FILE,
+  defaultTrack,
+  CLIPS_DIR,
+  eventLine,
+  formatTrack,
+  LECTURE_ID,
+  LECTURE_RENDERER,
+  LECTURES_DIR,
+  parseTrack,
+  PLAYER_RENDERER,
+  SCRIPT_TRACK,
+  sortEvents,
+  spokenDuration,
+  TRACK_ID,
+  TRACK_VERSION,
+  trackDuration,
+  utteranceLine,
+} from './track.ts'
+export type { AudioSegment, LectureTrack, MediaSegment, TrackBlock, Translation, Utterance, VideoSegment, Voice, Word } from './track.ts'
+export { isStale, translationDigest, translationItems } from './translation.ts'
+export type { TranslationItem } from './translation.ts'
+export { captionLanguages, captionsVtt, captionText, chaptersVtt, eventsVtt, vttTime } from './vtt.ts'
+export type { Captions } from './vtt.ts'
+export { charOffset, isWide, locate, realign, trackWords, wordIndexAt } from './words.ts'
+export { cueFor, narrationDraft, narrationSource } from './writeback.ts'
+export type { DraftSection, DraftStep, NarrationDraft, NarrationSource } from './writeback.ts'

package/src/domain/messages.ts

@@ -0,0 +1,324 @@
+import type { Locale } from '@lincyaw/kernel/contracts'
+
+export interface LectureMessages {
+  view: string
+  production: {
+    missingSpeech: string
+    cacheMissing(text: string, section: string, reason: string): string
+    generateHint: string
+    incompleteCache: string
+    noVoice: string
+    noPreparedLecture: string
+    /** Why a sentence has no voiceover, by the code a speech service or the plan gives; others read as `reasonGeneric`. */
+    reasons: Record<string, string>
+    reasonGeneric: string
+  }
+  /** The problems of a {cue} mark, written into the document's diagnostics. */
+  cue: {
+    problem(command: string, reason: string): string
+    unknown(name: string, commands: readonly string[]): string
+    pauseDuration(arg: string): string
+    clickAction: string
+    markName: string
+  }
+  recorder: {
+    region: string
+    record: string
+    rerecord: string
+    stop: string
+    camera: string
+    screen: string
+    track: string
+    newTrack: string
+    recording: (time: string) => string
+    saving: string
+    saved: (id: string) => string
+    failed: (reason: string) => string
+    unsupported: string
+    laser: string
+    marker: string
+    clear: string
+  }
+  player: {
+    playbackTimeout: string
+    region: string
+    play: string
+    pause: string
+    seek: string
+    time: (now: string, total: string) => string
+    rate: string
+    more: string
+    presenter: string
+    previous: string
+    next: string
+    captions: string
+    transcript: string
+    chapters: string
+    sideTabs: string
+    language: string
+    close: string
+    synthetic: string
+    silent: string
+    syntheticBy: (provider: string) => string
+    loading: string
+    failed: (reason: string) => string
+    missing: (id: string) => string
+    dropped: (count: number) => string
+    unsaid: (count: number) => string
+    stale: string
+    jump: (time: string) => string
+    step: (step: number) => string
+  }
+  list: {
+    title: string
+    empty: string
+    unavailable: string
+    open: (id: string) => string
+  }
+  editor: {
+    region: string
+    transcribe: (provider: string) => string
+    transcribeConfirm: (id: string, provider: string, local: boolean) => string
+    transcribing: string
+    transcribed: (count: number) => string
+    voice: string
+    synthesize: (provider: string) => string
+    synthesizeConfirm: (provider: string, voice: string, count: number, local: boolean) => string
+    synthesizing: string
+    synthesized: string
+    local: string
+    confirm: string
+    cancel: string
+    edit: (id: string) => string
+    save: string
+    saveSay: (provider: string) => string
+    saveSayConfirm: (provider: string, local: boolean) => string
+    saving: string
+    failed: (reason: string) => string
+  }
+}
+
+/** What the lecture's tools say to the AI and the author for the codes they fail with (the AI may pass it on). */
+export const LECTURE_TOOL_ERRORS: Record<Locale, Record<string, string>> = {
+  'zh-CN': {
+    'lecture-plugin-missing': '这份文稿没有启用讲授插件。',
+    'document-not-ready': '文稿还没有准备好，请稍后重试。',
+    'speech-unavailable': '未连接语音服务。请在文稿中启用语音服务插件 addon-speech。',
+    'track-required': '需要给出讲授的名称（track）。',
+    'lang-required': '需要给出要译成的语言（lang）。',
+  },
+  en: {
+    'lecture-plugin-missing': 'This document doesn’t use the lecture plugin.',
+    'document-not-ready': 'The document isn’t ready yet. Try again shortly.',
+    'speech-unavailable': 'No voice service is connected. Enable the voice service plugin addon-speech in the document.',
+    'track-required': 'Give the name of the lecture (track).',
+    'lang-required': 'Give the language to translate into (lang).',
+  },
+}
+
+export const LECTURE_MESSAGES: Record<Locale, LectureMessages> = {
+  'zh-CN': {
+    view: '讲授',
+    production: {
+      missingSpeech: '读取已生成的配音需要语音服务。请在文稿的 metadata.plugins 中启用语音服务插件 addon-speech。',
+      cacheMissing: (text, section, reason) => `页面 ${section} 的讲稿“${text}”没有与当前文字和音色一致的配音。${reason}`,
+      generateHint: '请先为当前讲稿和音色生成配音，再重新导出。检查和导出不会自动生成缺失的配音。',
+      incompleteCache: '当前讲稿与音色的配音不完整。请先生成配音，再导出。',
+      noVoice: '视频制作设置未指定音色，文稿没有生成的配音。',
+      noPreparedLecture: '文稿还没有可用的配音。请提供录音，或选择音色并生成配音。',
+      reasons: {
+        'voiceover-not-generated': '还没有为当前讲稿和音色生成配音。',
+        'voiceover-cache-missing': '语音服务没有给出已生成配音的文件。',
+        'speech-unavailable': '未连接语音服务。',
+        'voice-unset': '没有选择音色。',
+        'voice-step-binding-unavailable': '这句讲稿所在的步骤已经变化，请重新选择音色。',
+        'production-media-changed': '配音文件在读取时发生了变化，请重试。',
+      },
+      reasonGeneric: '无法读取这句讲稿的配音。',
+    },
+    cue: {
+      problem: (command, reason) => `提示“${command}”：${reason}`,
+      unknown: (name, commands) => `没有“${name}”这种提示。可用的提示：${commands.join('、')}。`,
+      pauseDuration: arg => `停顿需要时长，例如“pause 800ms”，而不是“${arg}”。`,
+      clickAction: '点击需要组件的操作名，例如“click reset”。',
+      markName: '标记需要名称，例如“mark merge”。',
+    },
+    recorder: {
+      region: '讲授录制',
+      record: '录制',
+      rerecord: '重录本页',
+      stop: '停止',
+      camera: '摄像头',
+      screen: '屏幕',
+      track: '讲授',
+      newTrack: '新建讲授',
+      recording: time => `正在录制 ${time}`,
+      saving: '正在保存…',
+      saved: id => `已保存 ${id}`,
+      failed: reason => `录制失败：${reason}`,
+      unsupported: '此浏览器无法录音',
+      laser: '激光笔',
+      marker: '荧光笔',
+      clear: '清除标记',
+    },
+    player: {
+      playbackTimeout: '无法开始播放。请重试。',
+      region: '讲授回放',
+      play: '播放',
+      pause: '暂停',
+      seek: '进度',
+      time: (now, total) => `${now} / ${total}`,
+      rate: '倍速',
+      more: '更多播放选项',
+      presenter: '演示者视图',
+      previous: '上一步',
+      next: '下一步',
+      captions: '字幕',
+      transcript: '字幕列表',
+      chapters: '章节',
+      sideTabs: '字幕列表与章节',
+      language: '字幕语言',
+      close: '关闭讲授',
+      synthetic: 'AI 生成的配音',
+      silent: '无声播放',
+      syntheticBy: provider => `AI 生成的配音（${provider}）`,
+      loading: '正在加载讲授…',
+      failed: reason => `无法加载讲授：${reason}`,
+      missing: id => `找不到讲授 ${id}`,
+      dropped: count => `文稿改动后有 ${count} 个操作失效，回放时略过`,
+      unsaid: count => `有 ${count} 句尚未生成配音，按估计的时长无声播放并显示字幕`,
+      stale: '译文需要更新：原句在翻译之后有改动',
+      jump: time => `跳到 ${time}`,
+      step: step => `第 ${step} 步`,
+    },
+    list: {
+      title: '讲授',
+      empty: '这份文稿还没有讲授',
+      unavailable: '无法读取讲授列表',
+      open: id => `播放讲授 ${id}`,
+    },
+    editor: {
+      region: '讲授的讲稿与配音',
+      transcribe: provider => `转写（${provider}）`,
+      transcribeConfirm: (id, provider, local) => (local ? `用 ${provider} 转写讲授 ${id} 的录音？` : `把讲授 ${id} 的录音发给 ${provider} 转写？转写由该服务计费。`),
+      transcribing: '正在转写…',
+      transcribed: count => `已转写 ${count} 句`,
+      voice: '音色',
+      synthesize: provider => `重新生成配音（${provider}）`,
+      synthesizeConfirm: (provider, voice, count, local) => `用 ${provider} 的音色“${voice}”为这段讲授的 ${count} 句话重新生成配音？${local ? '' : '生成由该服务计费，'}结果由 AI 生成。`,
+      synthesizing: '正在生成…',
+      synthesized: '已重新生成',
+      local: '本机，不外发',
+      confirm: '确定',
+      cancel: '取消',
+      edit: id => `编辑语句 ${id}`,
+      save: '保存',
+      saveSay: provider => `保存并重新生成配音（${provider}）`,
+      saveSayConfirm: (provider, local) => (local ? `用 ${provider} 为改过的这句话重新生成配音？结果由 AI 生成。` : `把改过的这句话发给 ${provider} 重新生成配音？生成由该服务计费，结果由 AI 生成。`),
+      saving: '正在保存…',
+      failed: reason => `无法完成：${reason}`,
+    },
+  },
+  'en': {
+    view: 'Lecture',
+    production: {
+      missingSpeech: 'Reading the generated voiceover needs the voice service. Enable the voice service plugin addon-speech in the document’s metadata.plugins.',
+      cacheMissing: (text, section, reason) => `The script “${text}” on page ${section} has no voiceover that matches its current text and voice. ${reason}`,
+      generateHint: 'Generate the voiceover for the current script and voice first, then export again. Checking and exporting never generate missing voiceover.',
+      incompleteCache: 'The voiceover for the current script and voice is incomplete. Generate the voiceover before exporting.',
+      noVoice: 'The video production settings name no voice, so the document has no generated voiceover.',
+      noPreparedLecture: 'The document has no voiceover yet. Provide a recording, or choose a voice and generate the voiceover first.',
+      reasons: {
+        'voiceover-not-generated': 'No voiceover is generated for the current script and voice yet.',
+        'voiceover-cache-missing': 'The voice service gave no file for the generated voiceover.',
+        'speech-unavailable': 'No voice service is connected.',
+        'voice-unset': 'No voice is chosen.',
+        'voice-step-binding-unavailable': 'The step of this sentence changed. Choose its voice again.',
+        'production-media-changed': 'The voiceover file changed while it was read. Try again.',
+      },
+      reasonGeneric: 'Couldn’t read the voiceover of this sentence.',
+    },
+    cue: {
+      problem: (command, reason) => `Cue “${command}”: ${reason}`,
+      unknown: (name, commands) => `There is no cue “${name}”. The cues are ${commands.join(', ')}.`,
+      pauseDuration: arg => `A pause needs a duration, such as “pause 800ms”, not “${arg}”.`,
+      clickAction: 'A click needs the action of the component, such as “click reset”.',
+      markName: 'A mark needs a name, such as “mark merge”.',
+    },
+    recorder: {
+      region: 'Lecture Recording',
+      record: 'Record',
+      rerecord: 'Re-record Page',
+      stop: 'Stop',
+      camera: 'Camera',
+      screen: 'Screen',
+      track: 'Lecture',
+      newTrack: 'New Lecture',
+      recording: time => `Recording ${time}`,
+      saving: 'Saving…',
+      saved: id => `Saved ${id}`,
+      failed: reason => `Recording failed: ${reason}`,
+      unsupported: 'This browser can’t record audio',
+      laser: 'Laser',
+      marker: 'Marker',
+      clear: 'Clear Marks',
+    },
+    player: {
+      playbackTimeout: 'Couldn’t start playback. Please try again.',
+      region: 'Lecture Playback',
+      play: 'Play',
+      pause: 'Pause',
+      seek: 'Position',
+      time: (now, total) => `${now} / ${total}`,
+      rate: 'Speed',
+      more: 'More Playback Options',
+      presenter: 'Presenter View',
+      previous: 'Previous Step',
+      next: 'Next Step',
+      captions: 'Captions',
+      transcript: 'Caption List',
+      chapters: 'Chapters',
+      sideTabs: 'Caption list and chapters',
+      language: 'Caption language',
+      close: 'Close Lecture',
+      synthetic: 'AI-generated voiceover',
+      silent: 'Silent playback',
+      syntheticBy: provider => `AI-generated voiceover (${provider})`,
+      loading: 'Loading the lecture…',
+      failed: reason => `Couldn’t load the lecture: ${reason}`,
+      missing: id => `Lecture ${id} not found`,
+      dropped: count => `${count} actions no longer match the document and are skipped`,
+      unsaid: count => `${count} sentences have no voiceover yet; they play silently for their estimated length, with captions`,
+      stale: 'Translation update needed: the sentence changed after it was translated',
+      jump: time => `Go to ${time}`,
+      step: step => `Step ${step}`,
+    },
+    list: {
+      title: 'Lectures',
+      empty: 'This document has no lecture yet',
+      unavailable: 'Couldn’t read the lectures',
+      open: id => `Play Lecture ${id}`,
+    },
+    editor: {
+      region: 'Lecture Script and Voiceover',
+      transcribe: provider => `Transcribe (${provider})`,
+      transcribeConfirm: (id, provider, local) => (local ? `Transcribe the audio of lecture ${id} with ${provider}?` : `Send the audio of lecture ${id} to ${provider} for transcription? The service charges for it.`),
+      transcribing: 'Transcribing…',
+      transcribed: count => `Transcribed ${count} sentences`,
+      voice: 'Voice',
+      synthesize: provider => `Regenerate Voiceover (${provider})`,
+      synthesizeConfirm: (provider, voice, count, local) => `Regenerate the voiceover for the ${count} sentences of this lecture in the voice “${voice}” of ${provider}? ${local ? 'The' : 'The service charges for it; the'} result is AI-generated.`,
+      synthesizing: 'Generating…',
+      synthesized: 'Regenerated',
+      local: 'On this machine; nothing is sent',
+      confirm: 'OK',
+      cancel: 'Cancel',
+      edit: id => `Edit Sentence ${id}`,
+      save: 'Save',
+      saveSay: provider => `Save and Regenerate Voiceover (${provider})`,
+      saveSayConfirm: (provider, local) => (local ? `Regenerate the voiceover for the corrected sentence with ${provider}? The result is AI-generated.` : `Send the corrected sentence to ${provider} to regenerate its voiceover? The service charges for it; the result is AI-generated.`),
+      saving: 'Saving…',
+      failed: reason => `Couldn’t finish: ${reason}`,
+    },
+  },
+}

package/src/domain/narration.ts

@@ -0,0 +1,160 @@
+import { deckLocale, isRecord, parseDuration } from '@lincyaw/kernel'
+import type { MdastNode, WordSpan } from '@lincyaw/kernel'
+import { LECTURE_MESSAGES } from './messages.ts'
+import { LECTURE_ID } from './track.ts'
+import { isWide } from './words.ts'
+
+export interface CueMark {
+  offset: number
+  command: string
+  to?: string
+  lines?: string
+  sel?: string
+}
+
+export interface SpokenText {
+  text: string
+  spoken: string
+  spans: WordSpan[]
+  cues: CueMark[]
+}
+
+const SILENT = new Set(['image', 'imageReference', 'html', 'footnoteReference', 'definition', 'footnoteDefinition', 'comment', 'mystComment', 'target', 'mystTarget'])
+
+const BLOCKS = new Set(['paragraph', 'listItem', 'heading', 'blockquote', 'tableCell', 'tableRow'])
+
+function text(value: unknown): string | undefined {
+  return typeof value === 'string' && value ? value : undefined
+}
+
+class Builder {
+  text = ''
+  spoken = ''
+  readonly spans: WordSpan[] = []
+  readonly cues: CueMark[] = []
+  private pending: 'space' | 'line' | undefined
+
+  append(shown: string, spoken = shown): void {
+    if (shown === spoken) {
+      for (const char of shown) this.char(char, char)
+      return
+    }
+    const [lead] = /^\s*/.exec(shown)!
+    for (const char of lead) this.char(char, char)
+    this.flush(shown.trim()[0] ?? '')
+    const start = this.text.length
+    const from = this.spoken.length
+    this.text += collapse(shown)
+    this.spoken += collapse(spoken)
+    this.spans.push({ from, to: this.spoken.length, start, end: this.text.length })
+    if (/\s$/.test(shown)) this.pending = 'space'
+  }
+
+  cue(mark: Omit<CueMark, 'offset'>): void {
+    this.cues.push({ ...mark, offset: this.text.length })
+  }
+
+  line(): void {
+    if (this.text) this.pending = 'line'
+  }
+
+  done(): SpokenText {
+    return { text: this.text, spoken: this.spoken, spans: this.spans, cues: this.cues.map(c => ({ ...c, offset: Math.min(c.offset, this.text.length) })) }
+  }
+
+  private char(shown: string, spoken: string): void {
+    if (/\s/.test(shown)) {
+      if (this.text) this.pending = shown === '\n' || this.pending === 'line' ? 'line' : 'space'
+      return
+    }
+    this.flush(shown)
+    this.text += shown
+    this.spoken += spoken
+  }
+
+  private flush(next: string): void {
+    const pending = this.pending
+    this.pending = undefined
+    if (!pending || !this.text) return
+    if (isWide(this.text.at(-1)) && (isWide(next) || !next)) return
+    this.text += ' '
+    this.spoken += ' '
+  }
+}
+
+function collapse(value: string): string {
+  return value.replace(/\s+/g, ' ').trim()
+}
+
+export function lectureRole(node: MdastNode, name: 'cue' | 'say'): boolean {
+  return node.type === 'presmiumRole' && node.plugin === LECTURE_ID && node.name === name
+}
+
+function walk(node: MdastNode, out: Builder): void {
+  if (SILENT.has(node.type)) return
+  if (node.type === 'presmiumRole') {
+    const options = isRecord(node.options) ? node.options : {}
+    if (lectureRole(node, 'cue')) {
+      out.cue({
+        command: (text(options.value) ?? '').trim(),
+        ...text(options.to) ? { to: text(options.to) } : {},
+        ...options.lines !== undefined && options.lines !== '' ? { lines: String(options.lines) } : {},
+        ...text(options.sel) ? { sel: text(options.sel) } : {},
+      })
+      return
+    }
+    if (lectureRole(node, 'say')) {
+      const shown = (node.children ?? []).map(child => plain(child)).join('') || (text(options.value) ?? '')
+      out.append(shown, text(options.alias) ?? shown)
+      return
+    }
+  }
+  if (node.type === 'break') {
+    out.line()
+    return
+  }
+  if (typeof node.value === 'string' && !node.children) {
+    out.append(node.value)
+    return
+  }
+  const block = BLOCKS.has(node.type)
+  if (block) out.line()
+  for (const child of node.children ?? []) walk(child, out)
+  if (block) out.line()
+}
+
+function plain(node: MdastNode): string {
+  if (SILENT.has(node.type)) return ''
+  if (typeof node.value === 'string' && !node.children) return node.value
+  return (node.children ?? []).map(plain).join('')
+}
+
+export function spokenText(node: MdastNode): SpokenText {
+  const out = new Builder()
+  walk(node, out)
+  return out.done()
+}
+
+export type CueCommand = 'next' | 'spot' | 'hint' | 'point' | 'click' | 'run' | 'ink' | 'pause' | 'mark'
+
+export const CUE_COMMANDS: readonly CueCommand[] = ['next', 'spot', 'hint', 'point', 'click', 'run', 'ink', 'pause', 'mark']
+
+export const TARGETED: ReadonlySet<CueCommand> = new Set(['spot', 'hint', 'point', 'click', 'run', 'ink'])
+
+/** What is wrong with a {cue} command, as a code; `cueProblem` words it in a document's language. */
+export type CueProblem = { code: 'unknown', name: string } | { code: 'pause-duration', arg: string } | { code: 'click-action' } | { code: 'mark-name' }
+
+export function parseCue(command: string): { name: CueCommand, arg: string } | { problem: CueProblem } {
+  const [name = '', ...rest] = command.trim().split(/\s+/)
+  const arg = rest.join(' ')
+  if (!(CUE_COMMANDS as readonly string[]).includes(name)) return { problem: { code: 'unknown', name } }
+  if (name === 'pause' && parseDuration(arg) === undefined) return { problem: { code: 'pause-duration', arg } }
+  if ((name === 'click' || name === 'mark') && !arg) return { problem: { code: name === 'click' ? 'click-action' : 'mark-name' } }
+  return { name: name as CueCommand, arg }
+}
+
+export function cueProblem(command: string, problem: CueProblem, lang: string | undefined): string {
+  const t = LECTURE_MESSAGES[deckLocale(lang)].cue
+  const reason = problem.code === 'unknown' ? t.unknown(problem.name, CUE_COMMANDS) : problem.code === 'pause-duration' ? t.pauseDuration(problem.arg) : problem.code === 'click-action' ? t.clickAction : t.markName
+  return t.problem(command, reason)
+}