/**
* VTT Pre-Processing Utility
*
* Strips all WebVTT technical metadata from transcript content,
* returning only clean speaker-attributed dialogue text.
*
* Removes:
* - "WEBVTT" header line
* - Cue timing lines (e.g., "00:00:00.000 --> 00:00:05.000")
* - Cue identifiers (numeric or UUID)
* - NOTE blocks
* - Blank/whitespace-only lines
* - HTML tags (<v>, <c>, etc.)
*/
/**
* Clean raw VTT transcript content into plain speaker dialogue.
*/
export function cleanVttTranscript(rawVtt: string): string {
const lines = rawVtt.split(/\r?\n/);
const cleanedLines: string[] = [];
let skipNote = false;
for (const line of lines) {
const trimmed = line.trim();
// Skip empty lines
if (trimmed === '') {
skipNote = false; // NOTE blocks end at a blank line
continue;
}
// Skip WEBVTT header
if (/^WEBVTT/i.test(trimmed)) {
continue;
}
// Skip NOTE blocks (multi-line comments)
if (/^NOTE\b/i.test(trimmed)) {
skipNote = true;
continue;
}
if (skipNote) {
continue;
}
// Skip timing lines: "00:00:00.000 --> 00:00:05.000"
if (/^\d{2}:\d{2}:\d{2}\.\d{3}\s*-->\s*\d{2}:\d{2}:\d{2}\.\d{3}/.test(trimmed)) {
continue;
}
// Skip pure numeric cue identifiers
if (/^\d+$/.test(trimmed)) {
continue;
}
// Skip UUID-style cue identifiers
if (/^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/i.test(trimmed)) {
continue;
}
// Strip HTML voice tags: <v Speaker Name>text</v> → Speaker Name: text
let cleaned = trimmed;
// Handle <v SpeakerName>dialogue</v> pattern (common in Teams VTT)
cleaned = cleaned.replace(/<v\s+([^>]+)>/gi, '$1: ');
cleaned = cleaned.replace(/<\/v>/gi, '');
// Strip any remaining HTML tags
cleaned = cleaned.replace(/<[^>]+>/g, '');
// Collapse multiple spaces
cleaned = cleaned.replace(/\s{2,}/g, ' ').trim();
if (cleaned.length > 0) {
cleanedLines.push(cleaned);
}
}
// Merge consecutive lines from the same speaker
return mergeSpeakerLines(cleanedLines);
}
/**
* Merge consecutive lines from the same speaker into single paragraphs.
* Input: ["Alice: Hello", "Alice: How are you", "Bob: Fine thanks"]
* Output: "Alice: Hello How are you\nBob: Fine thanks"
*/
function mergeSpeakerLines(lines: string[]): string {
if (lines.length === 0) return '';
const merged: string[] = [];
let currentSpeaker = '';
let currentText = '';
for (const line of lines) {
const speakerMatch = line.match(/^([^:]+):\s*(.*)/);
if (speakerMatch) {
const speaker = speakerMatch[1].trim();
const text = speakerMatch[2].trim();
if (speaker === currentSpeaker) {
// Same speaker → append
currentText += ' ' + text;
} else {
// New speaker → flush previous
if (currentSpeaker) {
merged.push(`${currentSpeaker}: ${currentText}`);
}
currentSpeaker = speaker;
currentText = text;
}
} else {
// No speaker prefix → append to current or add standalone
if (currentSpeaker) {
currentText += ' ' + line;
} else {
merged.push(line);
}
}
}
// Flush last speaker
if (currentSpeaker) {
merged.push(`${currentSpeaker}: ${currentText}`);
}
return merged.join('\n');
}
