Skip to content

KellisLab/auto-meeting-minutes

BranchesTags

Repository files navigation

Video Transcript Processing Pipeline

A comprehensive toolkit for processing video transcripts from Panopto URLs, converting subtitles to structured, searchable formats with AI-generated summaries. This platform transforms closed captions into rich, navigable documents with direct video timestamp links and intelligent content summaries.

Overview

This pipeline integrates several specialized Python scripts to create a complete workflow:

  1. url2id.py - Extract Panopto video ID from a URL
  2. url2meeting_name.py - Extract meeting name from Panopto URL for meaningful file naming
  3. url2file.py - Download SRT transcript from Panopto using video ID
  4. vtt2txt.py - Convert subtitles (VTT/SRT) to plain text transcripts with timestamps
  5. txt2xlsx.py - Convert text to Excel with speaker highlighting and visual formatting
  6. refineStartTimes.py - Improve timestamp matching for speakers with multiple appearances
  7. xlsx2html.py - Convert Excel to HTML with direct timestamp links and AI-generated summaries
  8. fullpipeline.py - Run the entire process from URL to HTML summaries in one command
  9. pipeline.py - Process local VTT/SRT files (alternative to URL-based processing)
  10. html_bold_converter.py - Convert markdown-style bold formatting to HTML bold tags

Setup and Installation

Prerequisites

  • Python 3.12 or higher
  • pip (Python package installer)
  • OpenAI API key (for AI-generated summaries)

Manual Installation

Linux/macOS

If you prefer a local installation on Linux or macOS:

  1. Clone the repository
git clone https://github.com/KellisLab/auto-meeting-minutes.git
cd auto-meeting-minutes
  1. Set up a Python virtual environment
python -m venv venv
source venv/bin/activate
  1. Install dependencies
pip install -r requirements.txt
  1. Download required NLTK data
python -c "import nltk; nltk.download('punkt'); nltk.download('stopwords'); nltk.download('punkt_tab')"
  1. Create a .env file with your configuration
echo "DJANGO_SECRET_KEY=your-secret-key-here" > .env
echo "DJANGO_DEBUG=True" >> .env
echo "DATABASE_URL=postgresql://username:password@host:port/database" >> .env
echo "PANOPTO_FOLDER_ID=your-panopto-folder-id" >> .env

Windows

For Windows users:

  1. Clone the repository
git clone https://github.com/KellisLab/auto-meeting-minutes.git
cd auto-meeting-minutes
  1. Set up a Python virtual environment
python -m venv venv
venv\Scripts\activate
  1. Install dependencies
pip install -r requirements.txt
  1. Download required NLTK data
python -c "import nltk; nltk.download('punkt'); nltk.download('stopwords'); nltk.download('punkt_tab')"
  1. Create a .env file with your configuration (using PowerShell)
# Create .env file with Django and database configuration
"DJANGO_SECRET_KEY = ""your-secret-key-here""" | Out-File -FilePath .env
"DJANGO_DEBUG = True" | Add-Content -Path .env
"DATABASE_URL = ""postgresql://username:password@host:port/database""" | Add-Content -Path .env
"PANOPTO_FOLDER_ID = ""your-panopto-folder-id""" | Add-Content -Path .env
  1. Run the application
# Run the web application
python app.py

# Or process a transcript directly
python fullpipeline.py https://mit.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=YOUR_VIDEO_ID

Web Interface Usage

The web interface provides a simple way to process Panopto video transcripts:

  1. Paste a Panopto video URL in the input field
  2. Configure options:
    • Skip timestamp refinement: Faster but less accurate topic mapping
    • HTML Format: Choose between numbered or simple format
    • Language: Select the transcript language
  3. Click "Process URL" and wait for processing to complete
  4. Download the generated files:
    • Original subtitle file (.srt)
    • Plain text transcript (.txt)
    • Excel spreadsheet (.xlsx)
    • Refined Excel with improved timestamps (.xlsx)
    • Speaker links with summaries (.html)
    • Meeting summaries (.html)
    • Speaker and meeting summaries (.md)

Command Line Usage

1. Full Pipeline (URL to Summaries)

Process a Panopto video URL from start to finish:

python fullpipeline.py https://mit.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=YOUR_VIDEO_ID [options]

Options:

  • --skip-refinement: Skip the start time refinement step
  • --language LANGUAGE: Language code (default: "English_USA")
  • --meeting-root DIRECTORY: Output directory for processed files
  • --skip-timestamps: Skip file timestamp adjustments
  • --enhanced-summaries: Use enhanced speaker summaries with multiple topics

2. Local File Processing

Process a local VTT or SRT file:

python pipeline.py input.srt VIDEO_ID [options]

Options:

  • --skip-refinement: Skip the start time refinement step
  • --html-format {simple|numbered}: Output format for HTML
  • --output-dir DIRECTORY: Directory to store output files
  • --meeting-name NAME: Custom name for output files

3. Individual Component Usage

Each script can also be used independently:

Extract Video ID from URL

python url2id.py [url]

Download SRT Transcript

python url2file.py [url] [output_file] [--language LANGUAGE]

Convert SRT/VTT to Text

python vtt2txt.py input.vtt [output.txt]

Convert Text to Excel

python txt2xlsx.py input.txt [output.xlsx]

Refine Timestamps

python refineStartTimes.py input.xlsx [output.xlsx]

Convert Excel to HTML with Summaries

python xlsx2html.py input.xlsx VIDEO_ID [output.html] [--format={simple|numbered}]

Output Files

For a video with ID meeting_name, the pipeline produces:

  • meeting_name.srt - Original subtitle file
  • meeting_name.txt - Plain text transcript with timestamps
  • meeting_name.xlsx - Excel formatted transcript with speaker highlighting
  • meeting_name_refined.xlsx - Excel with improved timestamp matching
  • meeting_name_speaker_summaries.html - HTML with speaker links and summaries
  • meeting_name_meeting_summaries.html - HTML with batch summaries and topic links
  • meeting_name_speaker_summaries.md - Markdown speaker summaries
  • meeting_name_meeting_summaries.md - Markdown meeting summaries

Advanced Features

Enhanced Speaker Summaries

The pipeline can generate detailed speaker-specific summaries with multiple topics per speaker:

  • Identifies distinct topics for each speaker using NLP
  • Creates separate summaries for each topic with timestamps
  • Links directly to the exact point in the video where each topic begins
  • Formats summaries in both HTML and Markdown formats

Enable with --enhanced-summaries flag or enable in xlsx2html.py.

Timestamp Refinement

The refineStartTimes.py script improves the accuracy of speaker timestamp links:

  • Analyzes text content for better topic-to-timestamp matching
  • Uses NLP to identify the most relevant speaker instances
  • Updates Excel with improved timestamp mappings
  • Generates corrected markdown files with accurate timestamps

Customization

Various parameters can be adjusted:

  • In txt2xlsx.py: Color generation, column formatting
  • In refineStartTimes.py:
    • Text similarity thresholds
    • Keyword extraction settings
    • Maximum time gap for matching
  • In xlsx2html.py:
    • Batch size for summaries (DEFAULT_BATCH_SIZE_MINUTES)
    • AI model selection (via .env file)
    • Summary prompt templates
    • HTML and Markdown formatting

Environment Variables

Create a .env file in the project directory to configure:

# Django Configuration
DJANGO_SECRET_KEY=your-secret-key-here
DJANGO_DEBUG=True

# Database Configuration
DATABASE_URL=postgresql://username:password@host:port/database

# Panopto Configuration
PANOPTO_FOLDER_ID=your-panopto-folder-id

# Celery Configuration (optional)
CELERY_BROKER_URL=redis://localhost:6379/0

Troubleshooting

Common Issues

  • Missing NLTK Data: If you get NLTK errors, run the NLTK download commands manually.
  • API Key Issues: Check that your API_KEY is correctly set in .env or as an environment variable.
  • Memory Errors: For large transcripts, try increasing Docker container memory allocation.
  • File Permission Issues: Ensure the temp_files directory is writable.

Logs

Check the application logs for detailed error information. In Docker:

docker logs transcript-processor

Project Structure

.
├── app.py                    # Flask web application
├── docker-compose.yml        # Docker Compose configuration
├── Dockerfile                # Docker build configuration
├── fullpipeline.py           # Full URL-to-summaries pipeline
├── pipeline.py               # Local file processing pipeline
├── html_bold_converter.py    # Convert markdown bold to HTML
├── refineStartTimes.py       # Timestamp refinement logic
├── requirements.txt          # Python dependencies
├── speaker_summary_utils.py  # Speaker summary generation utilities
├── templates/                # Flask HTML templates
│   └── index.html            # Web interface
├── url2file.py               # URL to SRT downloader
├── url2id.py                 # URL to video ID extractor
├── url2meeting_name.py       # URL to meeting name extractor
├── utils.py                  # Shared utility functions
├── vtt2txt.py                # VTT/SRT to text converter
└── xlsx2html.py              # Excel to HTML/MD converter

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues for improvements.

License

This project is available under the MIT License.

Acknowledgments

This pipeline was developed to help process and analyze video content efficiently, with a focus on academic and research meetings.

About

No description, website, or topics provided.

Resources

Readme
Activity
Custom properties

Stars

0 stars

Watchers

1 watching

Forks

2 forks
Report repository

Releases

No releases published

Packages

No packages published

Contributors 5

Languages