---
title: Profanity Detection & Replacement System
emoji: 🚫
colorFrom: red
colorTo: blue
sdk: gradio
sdk_version: 5.21.0
app_file: profanity_detector.py
pinned: false
---

# Profanity Detection & Replacement System

This app provides a comprehensive solution for detecting and cleaning profanity from both text and audio content. It uses state-of-the-art machine learning models to analyze content, identify inappropriate language, and generate clean alternatives.

### GitHub Repository

<img src="https://briantham.com/assets/img/projects/qr-code/Profanity-Detection-github-qr-code.svg?sanitize=true" alt="QR Code" width="300" />


## Features

- 🔍 Real-time profanity detection with adjustable sensitivity
- 🔄 Automatic text rephrasing to clean alternatives
- 🎤 Speech-to-text conversion with profanity filtering
- 🗣️ Text-to-speech generation for clean content
- 💻 User-friendly Gradio interface
- 🔄 Real-time streaming support for live audio processing

## Models Used

- Profanity Detection: `parsawar/profanity_model_3.1`
- Text Detoxification: `s-nlp/t5-paranmt-detox`
- Speech Recognition: OpenAI Whisper (large-v2)
- Text-to-Speech: Microsoft SpeechT5

## Requirements

- Python 3.10
- PyTorch with CUDA support
- Gradio
- Transformers
- OpenAI Whisper
- Other dependencies listed in `requirements.txt`

## Interface

The app provides three main interaction modes:

1. **Text Analysis**: Enter text to detect and clean profanity
2. **Audio Analysis**: Upload or record audio for profanity detection
3. **Real-time Streaming**: Process live audio with instant profanity filtering

## Technical Details

- GPU acceleration supported for faster processing
- Memory-optimized with FP16 precision where available
- Configurable profanity detection threshold
- Built-in error handling and logging
- Dark mode support

## Team Members

- Brian Tham
- Hong Ziyang
- Nabil Zafran
- Adrian Ian Wong
- Lin Xiang Hong