Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 | 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 15x 15x 8x 7x 15x 1x 1x 1x 1x 1x 20x 20x 20x 20x 1x 1x 1x 1x 1x 1x 3x 3x 3x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 3x 3x 2x 2x 2x 2x 2x 2x 8x 8x 8x 2x 2x 2x 2x 2x 2x 1x 2x 2x 2x 2x 2x 2x 1x 1x 1x 2x 5x 5x 5x 5x 5x 5x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 5x 5x 5x 5x 5x 5x 5x 5x 5x 5x 5x 5x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 1x 1x 2x 2x 10x 10x 10x 10x 10x 2x 2x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x | /**
* Word-alignment model and normalizer for ElevenLabs music generation.
*
* The detailed music endpoint returns a JSON blob with per-word timing data
* for the sung vocals; this module normalizes that raw JSON against the
* canonical lyrics we sent in the composition plan, producing a uniform
* `SyncedLyricsModel` the player can render without caring about ElevenLabs'
* exact response shape.
*
* Robustness goals:
* - Try several plausible shapes for the raw alignment (parallel arrays,
* array-of-objects). The exact field names from `/v1/music/detailed` aren't
* pinned by public docs at the time of writing — we lean on shape sniffing.
* - If alignment is missing or empty, fall back to a static-lyrics model with
* `words: null` per line so the UI can render plain text.
* - Zip canonical words to aligned words sequentially. The model sings what we
* asked it to sing, so ordering is reliable; if counts diverge, the extras
* on either side just lose timing.
*/
import type { MusicAlignmentJson } from '../elevenlabs/music-client'
// ============================================================================
// Public types
// ============================================================================
/** Subset of the composition plan the player needs to render lyrics. */
export interface SongLyricsSection {
name: string
lines: string[]
durationMs: number
/** Optional "behind the line" notes rendered under the section in the player. */
annotations?: string[]
}
export interface SyncedWord {
text: string
startMs: number
endMs: number
}
export interface SyncedLine {
/** Per-word timing when alignment matched; null = render line as static text. */
words: SyncedWord[] | null
/** Canonical line text — always populated, used for fallback rendering. */
rawText: string
startMs: number | null
endMs: number | null
}
export interface SyncedSection {
name: string
lines: SyncedLine[]
startMs: number | null
endMs: number | null
/** From the composition plan — used when alignment is absent. */
fallbackDurationMs: number
/** Optional "behind the line" notes — passed through from the source section. */
annotations?: string[]
}
export interface SyncedLyricsModel {
sections: SyncedSection[]
/** Best-effort total duration in ms. Prefers alignment end; falls back to plan. */
totalDurationMs: number
/** True when at least one line has per-word timing data. */
hasAlignment: boolean
}
// ============================================================================
// Helpers
// ============================================================================
interface FlatAlignedWord {
text: string
startMs: number
endMs: number
}
function asRecord(value: unknown): Record<string, unknown> | null {
return value && typeof value === 'object' && !Array.isArray(value)
? (value as Record<string, unknown>)
: null
}
/**
* Normalize a numeric timing field to ms. Heuristic: small values (< 1000)
* are assumed to be seconds; large values are assumed to be ms already.
*/
function toMs(value: number): number {
if (!Number.isFinite(value)) return 0
return value < 1000 ? value * 1000 : value
}
/**
* Split a line into words on whitespace. Punctuation stays attached to the
* word it borders ("hero," is one token) so the rendered line preserves
* the author's exact text.
*/
function tokenizeLine(line: string): string[] {
return line.split(/\s+/).filter((w) => w.length > 0)
}
// ============================================================================
// Raw-alignment extractor
// ============================================================================
/**
* Pull a flat list of `{text, startMs, endMs}` out of whatever ElevenLabs
* returned. Tries shapes in rough order of likelihood:
*
* 1. Parallel arrays at top level (TTS style):
* { words: [...], word_start_times_seconds: [...], word_end_times_seconds: [...] }
* 2. Array of objects at top level or under common keys:
* { words: [{ text|word, start|start_seconds|start_ms, end|... }] }
* 3. Nested under `alignment` / `song_metadata` / `metadata`.
*
* Returns [] when none match.
*/
function extractFlatWords(raw: MusicAlignmentJson | null): FlatAlignedWord[] {
if (!raw) return []
const candidates: Array<Record<string, unknown>> = []
const top = asRecord(raw)
if (top) {
candidates.push(top)
for (const key of ['alignment', 'song_metadata', 'songMetadata', 'metadata']) {
const nested = asRecord(top[key])
if (nested) candidates.push(nested)
}
}
for (const obj of candidates) {
const flat = tryParallelArrays(obj) ?? tryObjectArray(obj)
if (flat && flat.length > 0) return flat
}
return []
}
function tryParallelArrays(obj: Record<string, unknown>): FlatAlignedWord[] | null {
const words = obj.words
const starts =
obj.word_start_times_seconds ?? obj.word_start_times ?? obj.starts ?? obj.start_times
const ends = obj.word_end_times_seconds ?? obj.word_end_times ?? obj.ends ?? obj.end_times
if (!Array.isArray(words) || !Array.isArray(starts) || !Array.isArray(ends)) return null
const out: FlatAlignedWord[] = []
const n = Math.min(words.length, starts.length, ends.length)
for (let i = 0; i < n; i++) {
const text = words[i]
const start = starts[i]
const end = ends[i]
if (typeof text !== 'string' || typeof start !== 'number' || typeof end !== 'number') continue
out.push({ text, startMs: toMs(start), endMs: toMs(end) })
}
return out
}
function tryObjectArray(obj: Record<string, unknown>): FlatAlignedWord[] | null {
// Sweep likely keys for an array of word-objects. ElevenLabs `/v1/music/detailed`
// emits `words_timestamps` (with the plural 's' on words); other shapes are kept
// as defensive fallbacks in case the response format shifts.
const arrayKeys = ['words_timestamps', 'words', 'word_timestamps', 'lyrics', 'tokens']
for (const key of arrayKeys) {
const arr = obj[key]
if (!Array.isArray(arr)) continue
const out: FlatAlignedWord[] = []
for (const entry of arr) {
const e = asRecord(entry)
if (!e) continue
const text =
typeof e.text === 'string'
? e.text
: typeof e.word === 'string'
? e.word
: typeof e.token === 'string'
? e.token
: null
const startRaw =
e.start ?? e.start_time ?? e.start_seconds ?? e.startSeconds ?? e.start_ms ?? e.startMs
const endRaw = e.end ?? e.end_time ?? e.end_seconds ?? e.endSeconds ?? e.end_ms ?? e.endMs
if (!text || typeof startRaw !== 'number' || typeof endRaw !== 'number') continue
out.push({ text, startMs: toMs(startRaw), endMs: toMs(endRaw) })
}
if (out.length > 0) return out
}
return null
}
// ============================================================================
// Builder
// ============================================================================
/**
* Build a renderable lyrics model from canonical sections + raw alignment.
*
* Always returns a valid model; when alignment is missing the model still
* carries the lyrics as static text so the UI can render them.
*/
export function buildSyncedLyricsModel(
sections: SongLyricsSection[],
alignment: MusicAlignmentJson | null
): SyncedLyricsModel {
const flat = extractFlatWords(alignment)
const hasAlignment = flat.length > 0
let alignedIdx = 0
const syncedSections: SyncedSection[] = sections.map((section) => {
const lines: SyncedLine[] = section.lines.map((rawLine) => {
const wordTexts = tokenizeLine(rawLine)
if (!hasAlignment || wordTexts.length === 0) {
return { words: null, rawText: rawLine, startMs: null, endMs: null }
}
const syncedWords: SyncedWord[] = wordTexts.map((text) => {
if (alignedIdx >= flat.length) {
return { text, startMs: 0, endMs: 0 }
}
const a = flat[alignedIdx++]
// Keep the canonical text (preserves punctuation/casing the author wrote)
// but adopt the aligned timing.
return { text, startMs: a.startMs, endMs: a.endMs }
})
const startMs = syncedWords[0]?.startMs ?? null
const endMs = syncedWords[syncedWords.length - 1]?.endMs ?? null
return { words: syncedWords, rawText: rawLine, startMs, endMs }
})
const firstWithStart = lines.find((l) => l.startMs !== null)
const lastWithEnd = [...lines].reverse().find((l) => l.endMs !== null)
return {
name: section.name,
lines,
startMs: firstWithStart?.startMs ?? null,
endMs: lastWithEnd?.endMs ?? null,
fallbackDurationMs: section.durationMs,
annotations: section.annotations,
}
})
const totalFromAlignment = hasAlignment ? flat.reduce((max, w) => Math.max(max, w.endMs), 0) : 0
const totalFromPlan = sections.reduce((sum, s) => sum + s.durationMs, 0)
const totalDurationMs = totalFromAlignment > 0 ? totalFromAlignment : totalFromPlan
return {
sections: syncedSections,
totalDurationMs,
hasAlignment,
}
}
// ============================================================================
// Active-word lookup
// ============================================================================
export interface ActiveLyricLocation {
sectionIndex: number
lineIndex: number
/** -1 when the line has no per-word alignment (static fallback render). */
wordIndex: number
}
/**
* Find what's playing right now given the model and current ms.
*
* Picks the latest word whose `startMs <= currentMs`. Returns null when
* `currentMs` is before any aligned word, or when the model has no alignment.
*/
export function findActiveLocation(
model: SyncedLyricsModel,
currentMs: number
): ActiveLyricLocation | null {
if (!model.hasAlignment) return null
let last: ActiveLyricLocation | null = null
for (let si = 0; si < model.sections.length; si++) {
const section = model.sections[si]
for (let li = 0; li < section.lines.length; li++) {
const line = section.lines[li]
if (!line.words) continue
for (let wi = 0; wi < line.words.length; wi++) {
const w = line.words[wi]
if (currentMs >= w.startMs) {
last = { sectionIndex: si, lineIndex: li, wordIndex: wi }
} else {
return last
}
}
}
}
return last
}
|