Audio Processor

A Dockerized Python script that concatenates MP3 files, normalizes loudness, applies noise filtering using ffmpeg, and handles chapter markers. It also downloads metadata from Hardcover API.

Features

• Concatenates MP3 files from subdirectories
• Normalizes audio loudness using ffmpeg's loudnorm filter
• Applies noise reduction filter using ffmpeg's afftnoise filter
• Handles chapter markers (retrieves from source files or creates new ones)
• Downloads metadata from Hardcover API

Requirements

• Docker
• ffmpeg
• Python 3.9+

Installation

1. Clone this repository
2. Build the Docker image:

   docker build -t audio-processor .

Usage

docker run --rm \
  -v /path/to/input/directory:/input \
  -v /path/to/output/directory:/output \
  audio-processor \
  --input-dir /input \
  --output-file /output/output.mp3
  --book-id 12345

Arguments

• --input-dir: Directory containing MP3 files in subfolders (required)
• --output-file: Output file path (required)
• --book-id: Hardcover book ID for metadata (optional)
• --config-file: Path to config file (optional, defaults to config.yaml)

How It Works

1. Finding MP3 files: The script recursively searches the input directory for all MP3 files
2. Concatenation: Uses ffmpeg's concat demuxer to combine all MP3 files
3. Loudness Normalization: Applies ffmpeg's loudnorm filter to normalize audio loudness
4. Noise Filtering: Applies ffmpeg's afftnoise filter to reduce background noise
5. Chapter Marker Handling:
   • If source files contain chapter markers, they are preserved in the output
   • If source files don't contain chapter markers, new ones are created based on file durations
6. Metadata Retrieval: Attempts to fetch metadata from Hardcover API

Docker Volume Mounts

The script expects the following volume mounts:

• /input: Directory containing MP3 files
• /output: Directory for output files

Example

# Process audio files and save output to /output
docker run --rm \
  -v $PWD/input:/input \
  -v $PWD/output:/output \
  audio-processor \
  --input-dir /input \
  --output-file /output/combined.mp3

Configuration

The audio processor uses a config.yaml file to configure processing parameters. Here's an example configuration:

api:
  hardcover_api_key: "YOUR_HARDCOVER_API_KEY_HERE"

audio_processing:
  loudness_normalization:
    I: -16
    TP: -1.5
    LRA: 11
  noise_filter:
    noise_reduction: 0.5
    noise_floor: 0.3
    noise_profile: 0.2
  output:
    bitrate: "192k"

Notes

• The script creates temporary files during processing that are automatically cleaned up
• Audio normalization and noise filtering are applied to the concatenated audio file
• Hardcover API requires a valid API key to be configured in the config.yaml file
• Chapter markers are handled automatically, preserving existing markers or creating new ones when needed