NarrationTimeline

Builds and queries word-level timelines for recorded speech.

estimate is fallback scaffolding for a recording without measured alignment. Prefer NarrationAlignment whenever measured word timings are available.

Properties

val leadIn: Double = 0.35

Silence reserved before the first estimated word.

Time reserved for a breath at each line break in an estimated timeline.

fun estimate(passage: Passage, duration: Double, tokenizer: WordTokenizer, weighting: SpeechWeighting): List<WordTiming>

Estimates word timings by speech weight, reserving a pause at each line break.

Extends each supplied word interval through its release and the silence that follows.

Extends each word toward the next word, without passing it or exceeding limit.

Returns the index of the word sounding at at.

Returns the part of the recording occupied by one line.

Returns the part of the recording occupied by a consecutive run of lines.

Moves a line boundary out of speech and into the first resting silence in reach.

Returns the word sounding at at when it belongs to ofLines.

Returns the word at a fractional position through one line.

Returns the word at a fractional position through a consecutive run of lines.