fathom-extractor 1.0.0

Extract transcripts from HAR files, particularly from Fathom video calls

Fathom Extractor

A Python tool for extracting transcripts from HAR (HTTP Archive) files, particularly optimized for Fathom video call transcripts.

Features

• 🎥 Fathom Video Support: Specialized extraction for Fathom video call transcripts
• 🔍 Generic Transcript Detection: Finds transcripts from various APIs (Whisper, Deepgram, etc.)
• 📄 Multiple Output Formats: JSON, clean text, and beautiful Markdown with YAML frontmatter
• 🎯 Smart Pattern Matching: Automatically detects transcript-related network requests
• 📋 Rich Metadata: Extracts speakers, Q&A clips, AI notes, and meeting summaries
• ⚡ CLI Tool: Easy-to-use command-line interface

Installation

From PyPI (when published)

pip install fathom-extractor

From Source

git clone https://github.com/igutekunst/fathom-extractor.git
cd fathom-extractor
pip install -e .

Quick Start

1. Download a HAR file (see How to Download HAR Files)
2. Extract transcripts:

fathom-extractor recording.har

3. Get beautiful output:

fathom-extractor recording.har -m transcript.md -c clean.txt -v

Usage

Basic Usage

# Extract to JSON (default)
fathom-extractor recording.har

# Specify output file
fathom-extractor recording.har -o my_transcripts.json

# Create multiple output formats
fathom-extractor recording.har -m beautiful.md -c readable.txt

Command Line Options

fathom-extractor [-h] [-o OUTPUT] [-c CLEAN] [-m MARKDOWN] [-v] [--version] har_file

positional arguments:
  har_file              Path to the HAR file to extract transcripts from

options:
  -h, --help            show this help message and exit
  -o OUTPUT, --output OUTPUT
                        Output JSON file (default: extracted_transcripts.json)
  -c CLEAN, --clean CLEAN
                        Also create a clean, readable transcript file
  -m MARKDOWN, --markdown MARKDOWN
                        Create a beautiful markdown transcript with YAML frontmatter
  -v, --verbose         Enable verbose output
  --version             show program version and exit

Examples

# Basic extraction
fathom-extractor meeting.har

# Extract with all output formats and verbose logging
fathom-extractor meeting.har -o data.json -c transcript.txt -m report.md -v

# Just create a markdown report
fathom-extractor meeting.har -m meeting_notes.md

How to Download HAR Files

HAR (HTTP Archive) files capture all network traffic from your browser. Here's how to download them:

Chrome/Chromium

1. Open Developer Tools

   • Press F12 or Ctrl+Shift+I (Windows/Linux)
   • Press Cmd+Option+I (Mac)
   • Or right-click → "Inspect"

2. Go to Network Tab

   • Click the "Network" tab in Developer Tools
   • Make sure recording is enabled (red circle should be active)

3. Navigate and Capture

   • Go to your Fathom video page or transcript page
   • Let the page fully load and display the transcript
   • Scroll through the transcript if needed

4. Download HAR File

   • Right-click in the Network tab
   • Select "Save all as HAR with content"
   • Choose a filename and save

Firefox

1. Open Developer Tools

   • Press F12 or Ctrl+Shift+I (Windows/Linux)
   • Press Cmd+Option+I (Mac)

2. Go to Network Tab

   • Click the "Network" tab
   • Ensure recording is active

3. Capture Traffic

   • Navigate to your transcript page
   • Wait for full page load

4. Export HAR

   • Click the gear icon (⚙️) in the Network tab
   • Select "Save All As HAR"

Safari

1. Enable Developer Menu

   • Safari → Preferences → Advanced
   • Check "Show Develop menu in menu bar"

2. Open Web Inspector

   • Develop → Show Web Inspector
   • Go to Network tab

3. Capture and Export

   • Navigate to transcript page
   • Right-click in Network tab → "Export HAR"

Tips for Better Results

• Clear browser cache before recording to capture all requests
• Disable ad blockers temporarily to avoid missing requests
• Wait for full page load before saving the HAR file
• Interact with the page (scroll, click) to trigger all network requests
• For Fathom: Make sure you can see the full transcript on screen

Output Formats

JSON Output

Raw extracted data with full metadata and transcript content.

Clean Text Output

Human-readable format with:

• Meeting metadata
• Speaker information
• Q&A sections
• Full transcript with timestamps

Markdown Output

Beautiful formatted document with:

• YAML frontmatter with metadata
• Structured sections with emojis
• Proper formatting for speakers and timestamps
• Q&A sections with time ranges
• Meeting summaries and AI notes

What Gets Extracted

For Fathom Videos

• 👥 Speakers: Names and email addresses
• 📋 Meeting Summary: AI-generated meeting notes
• 💬 Q&A Clips: Questions and answers with timestamps
• 🤖 AI Notes: Additional AI-generated insights
• 📄 Full Transcript: Complete conversation with speaker attribution
• ⏰ Metadata: Meeting title, duration, host information

For Generic Transcripts

• 📝 Transcript Text: Raw or structured transcript data
• 🕒 Timestamps: When available
• 👤 Speaker Information: If present in the data
• 📊 Confidence Scores: From speech recognition APIs

Supported Sources

• Fathom Video: Full support for Fathom's transcript format
• OpenAI Whisper: API responses
• Deepgram: Transcript API responses
• Rev.ai: Speech-to-text API responses
• Google Speech-to-Text: API responses
• Azure Speech: API responses
• AWS Transcribe: API responses
• Generic APIs: Any API returning transcript-like JSON

Python API

You can also use the tool programmatically:

from fathom_extractor import HARTranscriptExtractor

# Create extractor
extractor = HARTranscriptExtractor('recording.har')

# Extract all transcripts
transcripts = extractor.extract_all_transcripts()

# Save in different formats
extractor.save_transcripts(transcripts, 'output.json')
extractor.create_clean_transcript(transcripts, 'clean.txt')
extractor.create_markdown_transcript(transcripts, 'beautiful.md')

# Access transcript data
for transcript in transcripts:
    print(f"Source: {transcript['source']}")
    print(f"URL: {transcript['url']}")
    if transcript['source'] == 'fathom':
        data = transcript['transcript_data']
        print(f"Speakers: {len(data.get('speakers', []))}")
        print(f"Q&A Clips: {len(data.get('qa_clips', []))}")

Troubleshooting

No Transcripts Found

If the tool doesn't find any transcripts:

1. Check the HAR file: Make sure you captured network traffic while viewing the transcript
2. Verify page loading: Ensure the transcript was fully loaded when you captured the HAR
3. Try verbose mode: Use -v flag to see what URLs were analyzed
4. Check browser: Some browsers or extensions might block certain requests

Incomplete Transcripts

If transcripts are missing content:

1. Scroll through the page: Some transcripts load content dynamically
2. Wait longer: Let the page fully load before capturing
3. Check network requests: Look for additional API calls in the Network tab

Large HAR Files

HAR files can be large. If you encounter memory issues:

1. Clear browser data before recording
2. Close other tabs to reduce network noise
3. Use incognito/private mode to avoid extension interference

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - see LICENSE file for details.

Author

Isaac Harrison Gutekunst

• GitHub: @igutekunst
• Email: [e-mail]

Changelog

v1.0.0

• Initial release
• Fathom video transcript extraction
• Generic transcript API support
• Multiple output formats
• CLI tool with comprehensive options