A comprehensive AI-powered tool for downloading, transcribing, and organizing the Voxology podcast. The goal is to create a searchable catalog that organizes episodes into series and enhances episode discovery using modern AI technologies.

💝 This project was lovingly made entirely with Claude Code for <$20 of API credits 🤖✨

The Voxology Catalog aims to:

• Download all episodes from the Voxology podcast website with metadata
• Transcribe audio content using AssemblyAI with speaker diarization
• Organize episodes by series using Gemini AI analysis
• Enhance discovery of related episodes and themes
• Analyze content statistics and costs

pip install -r requirements.txt

Required API Keys:

• AssemblyAI API key in ~/.ssh/assemblyai.txt
• Gemini API key in ~/.ssh/gemini_api_key.txt

# Run the complete end-to-end workflow
make all

Run these commands in order:

# 1. Scrape episodes (URLs, titles, dates) and extract audio links
make episodes

# 2. Download audio files
make download

# 3. Transcribe with AI speaker diarization
make transcribe

# 4. Analyze for series patterns
make series

# 5. Generate comprehensive statistics (optional)
make stats

# 6. Export data to CSV format (optional)
make csv

voxcatalog/
├── get_episode_links.py           # Scrapes episode URLs and metadata
├── get_episode_audio_links.py     # Extracts audio download links
├── get_audio_files.py             # Downloads audio files
├── transcribe-assemblyai.py       # AI transcription with speaker diarization
├── parse_series.py                # AI series analysis using Gemini
├── stats.py                       # Comprehensive statistics and analysis
├── parse_csv.py                   # CSV data export
├── episodes.json                  # Master episode database
├── series.json                    # Series organization data
├── stats.json                     # Detailed statistics
├── voxology_catalog.csv           # Spreadsheet-friendly data export
├── catalog/                       # Downloaded audio files and transcriptions
├── requirements.txt               # Python dependencies
├── Makefile                       # Command shortcuts
└── CLAUDE.md                      # Development guidance

{
  "episodes": [
    {
      "url": "https://www.voxologypodcast.com/episode-name/",
      "page": 1,
      "title": "Episode Title",
      "date": "2025-06-02",
      "audio_link": "https://media.../episode.mp3",
      "file_path": "catalog/episode-name.mp3",
      "transcription_file_path_assemblyai": "catalog/episode-name-assemblyai.txt",
      "audio_metadata": {
        "file_size_bytes": 15234567,
        "duration_seconds": 1825.4,
        "analyzed_date": "2025-01-02T15:30:45"
      }
    }
  ],
  "processed_pages": [1, 2, 3, ...],
  "last_updated": "2025-01-02T12:34:56"
}

{
  "INDEPENDENT": [
    "catalog/episode-1.mp3",
    "catalog/episode-5.mp3"
  ],
  "Exile Series": {
    "1": "catalog/exile-part-1.mp3", 
    "2": "catalog/exile-part-2.mp3"
  }
}

series_name,episode_num,episode_date,episode_url,episode_file_path_mp3
INDEPENDENT,1,2025-01-15,https://www.voxologypodcast.com/episode-1/,catalog/episode-1.mp3
Exile Series,1,2025-01-10,https://www.voxologypodcast.com/exile-part-1/,catalog/exile-part-1.mp3
Exile Series,2,2025-01-12,https://www.voxologypodcast.com/exile-part-2/,catalog/exile-part-2.mp3

• Resumable Processing: All scripts can be safely interrupted and resumed
• Progress Tracking: Real-time progress bars with tqdm
• Duplicate Detection: Skips already processed episodes and downloads
• Date Updates: Automatically updates missing episode dates
• Caching: Reuses audio metadata to speed up statistics

• AssemblyAI Transcription: Professional-quality speech-to-text with speaker diarization
• Gemini Series Analysis: AI-powered identification of podcast series patterns
• Conservative Classification: Only groups episodes into series when explicitly clear

• Audio Statistics: Duration, file sizes, storage requirements
• Transcription Analysis: Character counts, token estimates for LLM processing
• Cost Estimation: AssemblyAI transcription costs at $0.12/hour
• Series Organization: Automated categorization of episodes
• Failed Transcription Detection: Identifies episodes needing retry
• CSV Export: Spreadsheet-friendly data export for external analysis

Each script tracks its progress and can resume from where it left off:

• Episodes: Skips processed pages, updates missing dates automatically
• Audio Links: Skips episodes that already have audio_link field
• Downloads: Skips files that already exist, updates JSON progressively
• Transcriptions: Skips completed transcriptions, updates JSON per episode
• Series Analysis: Skips already analyzed episodes, saves progress continuously
• CSV Export: Generates fresh export from current data state

The Voxology podcast collection includes:

• 23 pages of episodes on the website
• 538 total episodes across all pages
• 100% episodes have dates (538/538)
• Average episode duration: ~60 minutes
• Total collection: ~535 hours of content
• Storage requirements: ~43GB for audio files
• Transcription progress: 14.5% complete (78/538 episodes)
• Series identification: 2 series discovered + independent episodes

Based on current collection:

• Total transcription cost: ~$64.23 (at $0.12/hour)
• Remaining transcription cost: ~$53.94 (460 episodes left)
• Average cost per episode: ~$0.12

Voxology is a podcast exploring theology, culture, and the intersection of faith and society. The catalog helps organize the extensive archive of episodes for easier navigation and discovery of related content and series.

See CLAUDE.md for development guidance and architectural notes.