Whisper Audiobook Chapterizer

This script is designed to split an audiobook into chapters based on detected chapter headings in the audio. It uses the extremely fast whisper.cpp model for transcription and ffmpeg for audio processing on Apple silion.

Features

• Transcribes audio files to text using the whisper.cpp model (default large-v3-turbo)
• Detects chapter headings in the transcribed text.
• Splits the audio file into separate chapters.
• Generates output files including SRT, CUE, Markdown, and raw text with timestamps.

System requirements

• Tested only on Mac os
• Procesor Apple silicon (M1/M2/M3)

Requirements

• Python 3.x
• ffmpeg
• mutagen
• pywhispercpp

Installation & usage

1. Clone the repository:
2. Install the required Python packages:

   pip install -r requirements.txt

3. Place your MP3 files in the Input/ directory.
4. Run the script:

   python Splitter.py

Customization

• Skip Phrases: You can customize the phrases to skip when detecting chapters by editing the skip_phrases.json file.

License

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

Acknowledgments

• whisper.cpp for the transcription model.
• FFmpeg for audio processing.