podangelex-JustAnotherCoderTheThird 1.0.2

Automatically remove profanity and toxic content from audio files using Whisper and Detoxify

PodangelEX

Automatically clean audio files by removing profanity and toxic content.

PodangelEX uses OpenAI's Whisper speech-to-text model combined with the Detoxify toxicity detector to automatically identify and remove profane, toxic, and harmful content from audio files.

Installation

Requirements: Python 3.8 or later, FFmpeg

pip install podangelex_JustAnotherCoderTheThird

System Requirements

• FFmpeg: Install via:
  • macOS: brew install ffmpeg
  • Ubuntu/Debian: sudo apt-get install ffmpeg
  • Windows: Download from ffmpeg.org

Quick Start

1. Install the package (see above)
2. Run the setup wizard:

   podangel
   

3. First run: The app will auto-create:
   • Configuration directory at ~/.podangelex/
   • Workspace at ~/podangelex_data/ with folders:
     • Input/ - Place audio files here
     • Output/ - Cleaned audio files appear here
     • .bridge/ - Temporary processing files
4. Add audio files to the Input/ folder
5. Run again: podangel to clean your files
6. Get results from the Output/ folder

How It Works

Step 1: Transcription

Whisper transcribes your audio file to text with word-level timestamps.

Step 2: Toxicity Detection

Two-phase approach:

• Exact matching: Checks transcribed words against a built-in profanity list
• Context-aware: Uses machine learning to detect toxic phrases even if not on the word list

Step 3: Audio Cutting

FFmpeg extracts only the clean portions of audio and concatenates them.

Configuration

On first run, you'll be asked to configure:

Model Size

• tiny (1GB) - Fastest, ~60% accuracy
• base (1GB) - Fast, ~70% accuracy
• small (2GB) - Balanced, ~75% accuracy (recommended)
• medium (5GB) - Better, ~80% accuracy
• large (10GB) - Best, ~85% accuracy
• turbo (6GB) - Latest, ~80% accuracy

Workers

Number of parallel files to process. Use more workers if you have lots of VRAM and many files.

Thresholds

Fine-tune what gets flagged as toxic (0-1 scale, higher = stricter):

• toxicity (t): General profanity
• severe_toxicity (st): Severe language
• obscene (o): Obscene content
• threat (th): Threats
• insult (i): Insults
• identity_attack (id): Slurs/hate speech

Environment Variables

Optional customization:

• PODANGELEX_HOME - Custom config directory (default: ~/.podangelex/)
• PODANGELEX_WORKSPACE - Custom workspace location (default: ~/podangelex_data/)

Troubleshooting

"ffmpeg: command not found"

Install FFmpeg using the commands above.

"ModuleNotFoundError: No module named 'whisper'"

Reinstall the package: pip install --upgrade podangelex_JustAnotherCoderTheThird

Audio not being cleaned properly

Adjust toxicity thresholds by re-running podangel and selecting option (1) to reconfigure.

License

CC0 1.0 Universal - Public Domain

AI Declaration

I used some AI to help debug the code, provide commit messages on Github, and to organize the files for package uploading