Transcript Synchronizer

A web application for merging timestamped VTT files with corrected TXT/PDF transcripts to produce accurate, timestamped transcriptions.

Problem Statement

When transcribing oral history interviews:

• Automated transcription (e.g., MacWhisper) produces VTT files with timestamps but uncorrected text
• Manual corrections are made in TXT or PDF format without timestamps
• Need to combine corrected text with original timestamps

Solution

This tool aligns the corrected transcript with the timestamped version and generates a new VTT file with:

• Corrected text from the TXT/PDF
• Accurate timestamps from the original VTT
• Timestamps only at speaker changes
• Automatic splitting of segments longer than 2 minutes

Features

• Client-side processing: All processing happens in your browser - no server needed, completely private
• Smart text alignment: Uses intelligent algorithms to match corrected text with timestamped versions
• Speaker detection: Automatically identifies and tracks speaker changes using labels (e.g., "John:", "Interviewer:")
• Smart segmentation: Timestamps only at speaker changes, with automatic splitting of segments longer than 2 minutes
• Drag-and-drop interface: Easy-to-use web interface

How to Use

1. Open the app: Simply open index.html in a modern web browser (Chrome, Firefox, Safari, Edge)

2. Upload your files:

   • Corrected Transcript (TXT or PDF): Your manually corrected transcript with speaker labels (TXT recommended for best results)
   • Timestamped VTT: The original VTT file from MacWhisper (or similar tool) with timestamps

3. Process: Click the "Process Files" button

4. Download: Download your new, corrected VTT file with accurate timestamps

Requirements

• Modern web browser with JavaScript enabled
• TXT or PDF with speaker labels in format: "Speaker Name:" or "INTERVIEWER:"
• VTT file with standard WebVTT format
• Recommendation: Export corrected transcripts as TXT from Adobe Acrobat Pro (File → Export To → Text) for cleanest results

Technical Details

Architecture

• Frontend: Pure HTML/CSS/JavaScript (no build process required)
• PDF Processing: PDF.js library for text extraction
• Text Alignment: Custom algorithm using similarity scoring and sequence matching
• Client-side only: No data leaves your computer

How It Works

1. Extracts text from the corrected TXT or PDF, identifying speaker segments
2. Parses the timestamped VTT file into structured data
3. Aligns the corrected text with the uncorrected text using similarity matching
4. Transfers timestamps from the original VTT to the corrected text
5. Merges consecutive segments from the same speaker
6. Splits segments longer than 2 minutes at natural boundaries
7. Generates a new VTT file ready for use

Development

To modify or extend the application:

# Clone the repository
git clone https://github.com/jeffpooley/transcript-synchronizer.git
cd transcript-synchronizer

# Open index.html in your browser
open index.html

File Structure

• index.html - Main application interface
• styles.css - Application styling
• app.js - Main application logic and UI handling
• transcript-parser.js - TXT/PDF text extraction and speaker detection
• vtt-parser.js - VTT file parsing and generation
• text-aligner.js - Text alignment and timestamp transfer algorithms

License

MIT License - feel free to use and modify for your projects.