audio_transcription_utils / transcript-tracer.js
ydshieh's picture
ydshieh HF staff
Add original transcript-tracer
2c4a6b5 verified
/*
© 2023 Samuel Bradshaw
https://github.com/samuelbradshaw/transcript-tracer-js
*/
// Global state variables
var ttIsInitialized = false;
var ttTranscripts;
var ttMediaPlayers;
var ttActivePlayer;
var ttLinkedDataByMediaUrl = {};
// Configuration variables (see README.md)
var ttBlockSelector = null;
var ttPhraseSelector = null;
var ttAlignmentFuzziness = 0;
var ttTimeOffset = 0;
var ttAutoScroll = null;
var ttClickable = false;
// Prepare transcripts and media players
function loadTranscriptTracer(options=null) {
if (document.readyState == 'loading') {
// Wait for document to load
document.addEventListener('DOMContentLoaded', function() {
loadTranscriptTracer(options);
});
return;
}
// Save user-provided options to configuration variables
if (options) {
if ('blockSelector' in options) ttBlockSelector = options.blockSelector;
if ('phraseSelector' in options) ttPhraseSelector = options.phraseSelector;
if ('alignmentFuzziness' in options) ttAlignmentFuzziness = options.alignmentFuzziness;
if ('timeOffset' in options) ttTimeOffset = options.timeOffset;
if ('autoScroll' in options) ttAutoScroll = options.autoScroll;
if ('clickable' in options) ttClickable = options.clickable;
}
// Reset Transcript Tracer to prevent unexpected behavior if loadTranscriptTracer() is called more than once
if (ttIsInitialized) {
if (ttTranscripts) for (const transcript of ttTranscripts) {
unlinkTranscript(transcript);
transcript.dataset.ttTranscript = '';
}
ttTranscripts = null;
ttMediaPlayers = null;
ttActivePlayer = null;
ttLinkedDataByMediaUrl = {};
document.querySelectorAll('.tt-word, .tt-whitespace').forEach(element => {
element.outerHTML = element.innerHTML;
});
}
// Set a few global state variables
ttIsInitialized = true;
ttTranscripts = document.getElementsByClassName('tt-transcript');
ttMediaPlayers = document.querySelectorAll('audio, video');
// Prepare transcript for alignment by adding spans and classes
for (let t = 0; t < ttTranscripts.length; t++) {
var transcript = ttTranscripts[t];
// Skip transcript if it's not linked to any media files
if (!transcript.dataset.ttMediaUrls) continue;
transcript.dataset.ttTranscript = t;
// Loop through text nodes to add spans (tt-word, tt-whitespace)
// Method of iterating text nodes thanks to https://stackoverflow.com/a/34700627/1349044
var iter = document.createNodeIterator(transcript, NodeFilter.SHOW_TEXT);
var textNode;
while (textNode = iter.nextNode()) {
const text = textNode.textContent;
if (text.replace(/\s/g, '').length != 0) {
var spannedText = '<span class="tt-word">' + text.replace(/(\s+)/g, '</span><span class="tt-whitespace">$1</span><span class="tt-word">') + '</span>';
spannedText = spannedText.replaceAll('<span class="tt-word"></span>', '');
// Replace text node with spanned text
const template = document.createElement('template');
template.innerHTML = spannedText;
textNode.parentNode.insertBefore(template.content, textNode);
textNode.parentNode.removeChild(textNode);
}
}
}
for (const mediaPlayer of ttMediaPlayers) {
// Link related transcript(s)
linkTranscripts(mediaPlayer);
// Add event listener to media player
mediaPlayer.addEventListener('play', function(e) {
if (ttActivePlayer != e.currentTarget) {
if (ttActivePlayer) {
ttActivePlayer.pause();
ttActivePlayer.removeEventListener('timeupdate', ttTimeUpdate);
}
ttActivePlayer = e.currentTarget;
if (ttCurrentTranscript) clearHighlightedWords(ttCurrentTranscript);
if (ttCurrentEvent) ttCurrentEvent = null;
}
ttActivePlayer.addEventListener('timeupdate', ttTimeUpdate);
var currentTranscript = ttTranscripts[ttLinkedDataByMediaUrl[ttActivePlayer.dataset.ttLinkedMediaUrl].transcriptIndex];
currentTranscript.dataset.ttCurrentMediaUrl = ttActivePlayer.dataset.ttLinkedMediaUrl;
});
mediaPlayer.addEventListener('ended', function(e) {
if (ttCurrentTranscript) clearHighlightedWords(ttCurrentTranscript);
if (ttCurrentEvent) ttCurrentEvent = null;
});
}
}
// Link media player to relevant transcripts
function linkTranscripts(mediaPlayer) {
var trackElement = mediaPlayer.querySelector('track[kind="metadata"]')
var mediaPlayerSourceUrls = [];
var mediaPlayerSrc = mediaPlayer.getAttribute('src');
var mediaPlayerSourceElements = mediaPlayer.querySelectorAll('source');
if (mediaPlayerSrc) mediaPlayerSourceUrls.push(mediaPlayerSrc);
if (mediaPlayerSourceElements) for (const s of mediaPlayerSourceElements) mediaPlayerSourceUrls.push(s.src);
// If there's nothing to link, return
if (!trackElement || !trackElement.getAttribute('src') || mediaPlayerSourceUrls.length == 0) return;
// Fetch WebVTT content and link related transcripts
for (const transcript of ttTranscripts) {
for (const mediaUrl of mediaPlayerSourceUrls) {
if (transcript.dataset.ttMediaUrls.includes(mediaUrl)) {
mediaPlayer.dataset.ttLinkedMediaUrl = mediaUrl;
unlinkTranscript(transcript);
fetch(trackElement.src)
.then(r => r.text())
.then(vttContent => linkTranscript(mediaPlayer, vttContent, transcript));
break;
}
}
}
function linkTranscript(mediaPlayer, vttContent, transcript) {
var wordTimings = parseVttToWordTimings(vttContent);
transcript.dataset.ttCurrentMediaUrl = mediaPlayer.dataset.ttLinkedMediaUrl;
function normalizedWord(word) {
// Convert to lowercase, normalize, and remove anything that's not a letter or number
return word.toLowerCase().normalize('NFD').replace(/[^\p{L}\p{N}]/gu, '');
}
// Add metadata to block and phrase containers (if ttBlockSelector and ttPhraseSelector are defined)
var blockContainers = ttBlockSelector ? transcript.querySelectorAll(ttBlockSelector) : [];
for (let c = 0; c < blockContainers.length; c++) blockContainers[c].dataset.ttBlock = c;
var phraseContainers = ttPhraseSelector ? transcript.querySelectorAll(ttPhraseSelector) : [];
for (let c = 0; c < phraseContainers.length; c++) phraseContainers[c].dataset.ttPhrase = c;
// Add metadata to each word span, and build timed events list
var timedEvents = [];
var wordTimingsIndex = 0;
var wordSpans = transcript.getElementsByClassName('tt-word');
for (let s = 0; s < wordSpans.length; s++) {
var span = wordSpans[s];
// Find the next word timing object that matches the current span's text
var initialWordTimingsIndex = wordTimingsIndex;
var maxFuzzyWordTimingsIndex = Math.min(wordTimingsIndex + ttAlignmentFuzziness, wordTimings.length - 1);
while (normalizedWord(span.innerText) != normalizedWord(wordTimings[wordTimingsIndex].text) && wordTimingsIndex <= maxFuzzyWordTimingsIndex) {
wordTimingsIndex += 1;
}
if (normalizedWord(span.innerText) != normalizedWord(wordTimings[wordTimingsIndex].text)) {
// Could not find matching word within the fuzziness range
wordTimingsIndex = initialWordTimingsIndex;
continue;
}
// Get the block, phrase, and word index
blockIndex = ttBlockSelector ? (span.closest(ttBlockSelector)?.dataset?.ttBlock ?? null) : wordTimings[wordTimingsIndex].blockIndex;
phraseIndex = ttPhraseSelector ? (span.closest(ttPhraseSelector)?.dataset?.ttPhrase ?? null) : wordTimings[wordTimingsIndex].phraseIndex;
wordIndex = wordTimings[wordTimingsIndex].wordIndex;
// Add block, phrase, and word index as metadata on the span
span.dataset.ttBlock = blockIndex;
span.dataset.ttPhrase = phraseIndex;
span.dataset.ttWord = wordIndex;
// Add timed event to timed events list
if (timedEvents.length != 0 && wordTimings[wordTimingsIndex].startSeconds == timedEvents[timedEvents.length-1].seconds) {
timedEvents[timedEvents.length-1].currentWordIndexes.push(wordIndex);
} else {
timedEvents.push({
'seconds': wordTimings[wordTimingsIndex].startSeconds,
'currentWordIndexes': [wordIndex],
'phraseIndex': phraseIndex,
'blockIndex': blockIndex,
});
}
wordTimingsIndex += 1;
}
// For a given element, find the first parent element containing relevant children
function findRelevantParent(startingElement, endingElement, childSelector, relevantChildSelector) {
var currentElement = startingElement;
while (currentElement && currentElement != endingElement) {
var currentElement = currentElement.parentElement;
var children = currentElement.querySelectorAll(childSelector);
var relevantChildren = document.querySelectorAll(relevantChildSelector);
if (children.length == relevantChildren.length) {
// Relevant parent found
return currentElement;
} else if (children.length > relevantChildren.length) {
// Failed to find a relevant parent
break;
}
}
return null;
}
// Add metadata to block and phrase containers (if ttBlockSelector and ttPhraseSelector aren't defined)
if (!ttBlockSelector) {
var count = wordTimings[wordTimings.length-1].blockIndex + 1;
for (let c = 0; c < count; c++) {
var startingElement = document.querySelector(`[data-tt-block="${c}"]`);
var blockContainer = findRelevantParent(startingElement, transcript, '[data-tt-word]', `[data-tt-word][data-tt-block="${c}"]`);
if (blockContainer) blockContainer.dataset.ttBlock = c;
}
}
if (!ttPhraseSelector) {
var count = wordTimings[wordTimings.length-1].phraseIndex + 1;
for (let c = 0; c < count; c++) {
var startingElement = document.querySelector(`[data-tt-phrase="${c}"]`);
var phraseContainer = findRelevantParent(startingElement, transcript, '[data-tt-word]', `[data-tt-word][data-tt-phrase="${c}"]`);
if (phraseContainer) phraseContainer.dataset.ttPhrase = c;
}
}
// Sort timed events list by time
timedEvents = timedEvents.sort(function(a, b) {
return a.seconds - b.seconds;
})
// Add reference data to ttLinkedDataByMediaUrl
var transcriptIndex = parseInt(transcript.dataset.ttTranscript);
ttLinkedDataByMediaUrl[mediaPlayer.dataset.ttLinkedMediaUrl] = {
'transcriptIndex': transcriptIndex,
'wordTimings': wordTimings,
'timedEvents': timedEvents,
'mediaElement': mediaPlayer,
'textTrackData': mediaPlayer.textTracks[0],
}
// Add click listeners to words
if (ttClickable) {
for (const word of document.querySelectorAll('.tt-word')) {
word.addEventListener('click', handleWordClick);
}
}
}
}
// Unlink transcript from previous VTT
function unlinkTranscript(transcript) {
clearHighlightedWords(transcript);
var ttLinkedElements = transcript.querySelectorAll('[data-tt-word]');
for (const element of ttLinkedElements) {
element.dataset.ttWord = '';
element.dataset.ttPhrase = '';
element.dataset.ttBlock = '';
}
var mediaUrl = transcript.dataset.ttCurrentMediaUrl;
if (mediaUrl) {
delete ttLinkedDataByMediaUrl[mediaUrl]
transcript.dataset.ttCurrentMediaUrl = '';
}
for (const word of document.querySelectorAll('.tt-word')) {
word.removeEventListener('click', handleWordClick);
}
}
// Convert WebVTT into a wordTimings list
function parseVttToWordTimings(vttContent) {
var wordTimings = [];
// Split WebVTT at each double line break
var vttSegments = vttContent.split(/\r?\n\r?\n/);
if (vttSegments.length == 0 || !vttSegments[0].startsWith('WEBVTT')) {
console.error('Error: Invalid VTT file. See https://developer.mozilla.org/en-US/docs/Web/API/WebVTT_API');
return;
}
// Loop through each segment of the WebVTT
// WebVTT documentation: https://developer.mozilla.org/en-US/docs/Web/API/WebVTT_API
var cueWordCounter = 0;
var cuePhraseCounter = 0;
var cueBlockCounter = 0;
for (let s = 0; s < vttSegments.length; s++) {
var segment = vttSegments[s];
// Skip segment if it's not a cue
if (segment.startsWith('WEBVTT') || /^STYLE\s/.test(segment) || /^NOTE\s/.test(segment)) continue;
// Convert VTT timestamps to seconds
function convertVttTimestampToSeconds(vttTimestamp) {
var parts = vttTimestamp.trim().split(':');
var seconds = 0;
if (parts.length == 3) { // 00:02:01.003
seconds = (parseFloat(parts[0]) * 60 * 60) + (parseFloat(parts[1]) * 60) + parseFloat(parts[2])
} else { // 02:01.003
seconds = (parseFloat(parts[0]) * 60) + parseFloat(parts[1])
}
return seconds;
}
// Parse cue
var cueIdentifier = cuePayload = '';
var cueStartSeconds = cueEndSeconds = null;
var segmentLines = segment.split(/\r?\n/);
for (const line of segmentLines) {
if (line.includes('-->')) {
// Cue timings (single line)
cueStartSeconds = wordStartSeconds = convertVttTimestampToSeconds(line.split('-->')[0].trim());
cueEndSeconds = convertVttTimestampToSeconds(line.split('-->')[1].trim());
} else if (cueStartSeconds == null) {
// Cue identifier (optional)
cueIdentifier += line;
} else {
// Cue payload (may be multiple lines)
var adjustedLine = line.replace(/(<\d*:?\d+:\d+\.\d+>)\s/g, ' $1').replace(/(<\d*:?\d+:\d+\.\d+>)(<\d*:?\d+:\d+\.\d+>)/g, '$1 $2');
cuePayload += adjustedLine;
var cueWords = adjustedLine.split(/\s/)
for (let word of cueWords) {
var matches = Array.from(word.matchAll(/(^<\d*:?\d+:\d+\.\d+>)?(.+$)/g));
if (matches.length == 1 && matches[0][1] != null) {
match = matches[0];
wordStartSeconds = convertVttTimestampToSeconds(match[1].replace('<', '').replace('>', ''));
word = match[2];
}
// Push word to wordTimings list
if (word) {
wordTimings.push({
'text': word,
'startSeconds': wordStartSeconds,
'endSeconds': cueEndSeconds,
'wordIndex': cueWordCounter,
'phraseIndex': cuePhraseCounter,
'blockIndex': cueBlockCounter,
})
cueWordCounter += 1;
}
}
cuePhraseCounter += 1;
}
}
cueBlockCounter += 1;
}
return wordTimings;
}
// Respond to timeupdate event (progress as the audio or video is playing)
var ttCurrentTranscript = null;
var ttPreviousEvent = null;
var ttCurrentEvent = null;
var ttNextEvent = null;
function ttTimeUpdate(e) {
// If the current player isn't active or doesn't have data, return
if (!ttActivePlayer || e.currentTarget != ttActivePlayer || !(ttActivePlayer.dataset.ttLinkedMediaUrl in ttLinkedDataByMediaUrl)) return;
var adjustedCurrentTime = ttActivePlayer.currentTime + (ttTimeOffset * -1);
var ttData = ttLinkedDataByMediaUrl[ttActivePlayer.dataset.ttLinkedMediaUrl];
// Make sure the correct transcript is selected
if (!ttCurrentTranscript || ttCurrentTranscript.dataset.ttTranscript != ttData.transcriptIndex) {
ttCurrentTranscript = document.querySelector(`[data-tt-transcript="${ttData.transcriptIndex}"]`);
}
// If before the first event, after the last event, or within the range of the current event, return
if (ttCurrentEvent && (ttCurrentEvent.seconds < ttData.timedEvents[0].seconds || ttCurrentEvent.seconds > ttData.timedEvents[ttData.timedEvents.length-1].seconds)) return;
if (ttCurrentEvent && ttNextEvent && ttCurrentEvent.seconds <= adjustedCurrentTime && ttNextEvent.seconds > adjustedCurrentTime) return;
// Clear words that were highlighted from the previous event
clearHighlightedWords(ttCurrentTranscript);
// Add highlights for the current event
for (let t = 0; t < ttData.timedEvents.length; t++) {
if (ttData.timedEvents[t].seconds <= adjustedCurrentTime && (!ttData.timedEvents[t+1] || adjustedCurrentTime < ttData.timedEvents[t+1]?.seconds)) {
ttPreviousEvent = ttData.timedEvents[t-1] || null;
ttCurrentEvent = ttData.timedEvents[t];
ttNextEvent = ttData.timedEvents[t+1] || null;
// Mark blocks
if (ttCurrentEvent.blockIndex != null) {
var blockElements = ttCurrentTranscript.querySelectorAll(`[data-tt-block="${ttCurrentEvent.blockIndex}"]`);
for (let b = 0; b < blockElements.length; b++) blockElements[b].classList.add((b==0 && !blockElements[b].classList.contains('tt-word')) ? 'tt-current-block-container' : 'tt-current-block');
}
// Mark phrases
if (ttCurrentEvent.phraseIndex != null) {
var phraseElements = ttCurrentTranscript.querySelectorAll(`[data-tt-phrase="${ttCurrentEvent.phraseIndex}"]`);
for (let p = 0; p < phraseElements.length; p++) phraseElements[p].classList.add((p==0 && !phraseElements[p].classList.contains('tt-word')) ? 'tt-current-phrase-container' : 'tt-current-phrase');
}
// Mark words
if (ttCurrentEvent.currentWordIndexes.length > 0) {
for (const wordIndex of ttCurrentEvent.currentWordIndexes) {
var wordElements = ttCurrentTranscript.querySelectorAll(`[data-tt-word="${wordIndex}"]`);
for (const wordElement of wordElements) wordElement.classList.add('tt-current-word');
}
for (const wordElement of ttCurrentTranscript.getElementsByClassName('tt-word')) {
if (wordElement.classList.contains('tt-current-word')) break;
wordElement.classList.add('tt-previous-word');
}
}
// Auto-scroll to the highlighted text
if (ttAutoScroll) {
var scrollOptions = { behavior: 'smooth', block: 'start', inline: 'nearest' }
if (ttAutoScroll == 'block' && ttPreviousEvent?.blockIndex != ttCurrentEvent.blockIndex) {
document.querySelector('.tt-current-block-container').scrollIntoView(scrollOptions);
} else if (ttAutoScroll == 'phrase' && ttPreviousEvent?.phraseIndex != ttCurrentEvent.phraseIndex) {
document.querySelector('.tt-current-phrase-container').scrollIntoView(scrollOptions);
} else if (ttAutoScroll == 'word') {
document.querySelector('.tt-current-word').scrollIntoView(scrollOptions);
}
}
break;
}
}
}
// Clear highlighted words in transcript
function clearHighlightedWords(transcript) {
if (!transcript) return;
var ttHighlightedElements = transcript.querySelectorAll('[class*="tt-current"], [class*="tt-previous"]');
for (const element of ttHighlightedElements) {
element.classList.remove('tt-current-block', 'tt-current-block-container', 'tt-current-phrase', 'tt-current-phrase-container', 'tt-current-word', 'tt-previous-word');
}
}
// Handle when a word in the transcript with an event listener is clicked
function handleWordClick(e) {
var wordElement = e.currentTarget;
var wordIndex = wordElement.dataset.ttWord;
var transcript = wordElement.closest('.tt-transcript');
var mediaUrl = transcript.dataset.ttCurrentMediaUrl;
var startSeconds = ttLinkedDataByMediaUrl[mediaUrl].wordTimings[wordIndex].startSeconds
ttLinkedDataByMediaUrl[mediaUrl].mediaElement.currentTime = startSeconds;
}