Podcast Teaser Generator

Automatically generate engaging audio teasers from your podcast episodes.

Overview

This tool analyzes your podcast audio tracks and automatically extracts the most interesting segments to create short, compelling teasers. It uses advanced audio analysis to identify segments with:

• High energy/excitement (volume peaks)
• Dynamic tonal variation (spectral contrast)
• Interesting speech rhythm patterns
• Clean cut points at natural breaks

Features

• Fully automated teaser generation
• Customizable teaser duration and settings
• Smart segment selection based on audio features
• Clean transitions with automatic crossfades
• Audio level normalization
• Optional visualization of audio analysis
• Batch processing of multiple files

Installation

1. Clone this repository:

git clone https://github.com/yourusername/podcast_teaser.git
cd podcast_teaser

2. Install dependencies:

pip install -r requirements.txt

3. Or install the package:

pip install -e .

Usage

Basic Usage

Generate a 60-second teaser from a podcast episode:

python podcast_teaser.py path/to/your/podcast.mp3

Advanced Options

python podcast_teaser.py path/to/your/podcast.mp3 --duration 90 --visualize --output-dir custom_output

Process all podcast files in a directory:

python podcast_teaser.py path/to/podcast/directory --config custom_config.json

Configuration

Customize the teaser generation by editing config.json or providing your own configuration file. The application uses the following configuration parameters:

Basic Teaser Settings

{
  "teaser_duration": 60,            // Target duration in seconds for individual teasers
  "segment_min_duration": 3,        // Minimum duration for each extracted segment in seconds
  "segment_max_duration": 15,       // Maximum duration for each extracted segment in seconds
  "num_segments": 5,                // Target number of segments to extract for each teaser
  "crossfade_duration": 500,        // Duration of crossfade between segments in milliseconds
  "output_format": "mp3",           // Audio format for output files (mp3, wav, etc.)
  "normalize_audio": true,          // Whether to normalize audio levels in final teaser

  // Analysis weights - controls how segments are selected
  "energy_weight": 0.4,             // Weight for energy-based detection (louder/excited moments)
  "spectral_weight": 0.3,           // Weight for spectral contrast (tonal variation)
  "tempo_weight": 0.2,              // Weight for speech tempo variations
  "silence_threshold": -40,         // dB threshold for silence detection

  // Intro/Outro Handling
  "exclude_intro_outro": true,      // Whether to exclude podcast intro/outro music
  "intro_duration": 30,             // Estimated duration of intro in seconds
  "outro_duration": 30,             // Estimated duration of outro in seconds

  // Summary Teaser Settings
  "create_summary_teaser": true,    // Whether to create a summary teaser when processing multiple files
  "summary_segments_per_track": 2,  // Number of segments to include per track in summary
  "summary_teaser_duration": 120,   // Target duration for summary teaser in seconds

  // Visualization
  "visualize": false                // Whether to generate visualization of audio analysis
}

Configuration Details

Basic Teaser Settings

• teaser_duration: Controls how long the generated teaser will be. Shorter teasers will be more selective.
• segment_min/max_duration: Limits how short or long extracted segments can be. Adjust based on your content.
• num_segments: Target number of highlights to extract. The actual number may be less if not enough quality segments are found.
• crossfade_duration: Longer values create smoother transitions but may cut into content.

Analysis Weights

• energy_weight: Higher values favor louder, more excited moments.
• spectral_weight: Higher values favor segments with more vocal variation (questions, tonal shifts).
• tempo_weight: Higher values favor segments with changing speech rhythms.
• Weights should sum to approximately 1.0 for best results.

Intro/Outro Handling

• When exclude_intro_outro is enabled, the specified durations at the beginning and end of episodes are ignored during analysis.
• Adjust intro_duration and outro_duration to match your podcast format.

Summary Teaser

• When processing multiple episodes, a summary teaser combines the best moments from all episodes.
• summary_segments_per_track controls how many segments from each episode are considered.
• This feature is ideal for creating "best of" compilations from multiple episodes.

How It Works

1. Audio Analysis: The system analyzes your podcast using multiple audio features:

   • RMS energy (volume/excitement)
   • Spectral contrast (tonal variation)
   • Speech tempo and rhythm patterns
   • Silence detection for natural breaks

2. Segment Selection: The most interesting moments are identified and scored.

3. Smart Extraction: Segments are intelligently extracted with clean cut points at natural breaks.

4. Teaser Assembly: Selected segments are combined with smooth crossfades.

5. Output: The final teaser is saved with optional visualization of the analysis.

Example Visualization

When using the --visualize option, the system generates plots showing:

• Combined interest score with highlighted selected segments
• RMS energy analysis
• Spectral contrast analysis
• Tempo/rhythm analysis

Tips for Better Results

• Longer episodes provide more material for selection
• Episodes with dynamic conversations tend to produce better teasers
• Try adjusting weights in config.json if teasers aren't capturing the right moments
• For interview podcasts, increasing the spectral_weight can help catch back-and-forth dialogue

Documentation

Full documentation is available at Read the Docs.

Contributing

Contributions are welcome! See CONTRIBUTING.md for details on how to contribute to this project.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Credits

Developed with ❤️ for podcast creators everywhere.