Accent Analyzer

A tool for analyzing English accents from video URLs for hiring purposes.

Features

• Accept a video URL (YouTube, Loom, or direct MP4 link)
• Extract audio from the video
• Detect if the speech is in English
• Classify the speaker's accent
• Provide a confidence score
• Generate an explanation of the analysis

Prerequisites & Installation

System Requirements

• Python 3.8 or higher (tested with Python 3.9-3.12)
• FFmpeg (required for audio extraction)
• Internet connection (for downloading videos and ML models)

Step 1: Install FFmpeg

FFmpeg is essential for audio extraction. Choose one method:

Windows:

Option A - Using Conda (Recommended):

conda install ffmpeg

Option B - Manual Installation:

1. Download FFmpeg from: https://www.gyan.dev/ffmpeg/builds/
2. Download "release builds" → "ffmpeg-release-essentials.zip"
3. Extract to C:\ffmpeg
4. Add C:\ffmpeg\bin to your Windows PATH:
   • Press Win+R, type sysdm.cpl
   • Go to "Advanced" → "Environment Variables"
   • Under "System Variables", find "Path" and click "Edit"
   • Click "New" and add C:\ffmpeg\bin
   • Click OK and restart your terminal/VS Code

Option C - Using Chocolatey:

choco install ffmpeg

macOS:

brew install ffmpeg

Ubuntu/Debian:

sudo apt-get update
sudo apt-get install ffmpeg

Verify FFmpeg Installation:

ffmpeg -version

You should see version information if installed correctly.

Step 2: Install Python Dependencies

Option A - Using pip (Standard):

pip install -r requirements.txt

Option B - Individual Installation (if requirements.txt fails):

# Core dependencies
pip install streamlit>=1.32.0
pip install numpy>=1.26.0
pip install librosa>=0.10.1
pip install torch>=2.1.0
pip install transformers>=4.35.0
pip install pydub>=0.25.1
pip install requests>=2.31.0
pip install soundfile>=0.12.1
pip install yt-dlp>=2023.12.30
pip install ffmpeg-python>=0.2.0

Option C - Using Virtual Environment (Recommended):

# Create virtual environment
python -m venv accent-analyzer-env

# Activate virtual environment
# Windows:
accent-analyzer-env\Scripts\activate
# macOS/Linux:
source accent-analyzer-env/bin/activate

# Upgrade pip and install dependencies
pip install --upgrade pip setuptools wheel
pip install -r requirements.txt

Step 3: Verify Installation

Test that all components work:

# Test Python imports
python -c "import streamlit; print('✓ Streamlit installed')"
python -c "import librosa; print('✓ Librosa installed')"
python -c "import torch; print('✓ PyTorch installed')"
python -c "import yt_dlp; print('✓ yt-dlp installed')"

# Test FFmpeg
ffmpeg -version

Running the Application

Local Development:

streamlit run app.py

The app will open in your browser at http://localhost:8501

Production Deployment:

Streamlit Cloud (Free):

1. Fork this repository to your GitHub account
2. Go to Streamlit Cloud
3. Sign in with your GitHub account
4. Deploy the app by selecting this repository
5. The app will be deployed at a public URL

Docker Deployment:

# Build the Docker image
docker build -t accent-analyzer .

# Run the container
docker run -p 8501:8501 accent-analyzer

Heroku Deployment:

1. Install Heroku CLI
2. Create a Procfile with: web: streamlit run app.py --server.port=$PORT --server.address=0.0.0.0
3. Deploy:

heroku create your-app-name
git push heroku main

How to Use

1. Start the application using one of the methods above
2. Enter a video URL in the input field:
   • YouTube: https://www.youtube.com/watch?v=VIDEO_ID
   • Loom: https://www.loom.com/share/VIDEO_ID
   • Direct MP4: Any direct link to an MP4 file
3. Click "Analyze Accent"
4. Wait for processing (may take 30-60 seconds for longer videos)
5. View results including accent classification, confidence score, and explanation

Supported Video Sources

• YouTube (public videos only)
• Loom (public recordings)
• Direct video links (.mp4, .avi, .mov, .mkv)
• Other platforms that provide direct video URLs

Troubleshooting

Common Issues:

"FFmpeg not found" error:

• Ensure FFmpeg is installed and in your PATH
• Restart your terminal/IDE after installation
• Try running ffmpeg -version to verify

"Error downloading YouTube video":

• Ensure the video is public and accessible
• Some videos may be region-restricted
• Try with a different video URL

Package installation errors:

• Update pip: pip install --upgrade pip
• Use a virtual environment
• Install packages individually if bulk installation fails

"Module not found" errors:

• Ensure all dependencies are installed
• Check if you're using the correct Python environment
• Reinstall requirements: pip install -r requirements.txt --force-reinstall

Performance Tips:

• Shorter videos process faster (under 5 minutes recommended)
• Good audio quality improves accent detection accuracy
• Clear speech without background music works best
• Single speaker videos give more accurate results

Technical Architecture

How It Works:

1. Video Download: Uses yt-dlp for YouTube or direct download for other URLs
2. Audio Extraction: FFmpeg extracts audio track from video
3. Audio Processing: Librosa processes audio and extracts acoustic features
4. Speech Recognition: Transcribes audio to text (framework ready for Whisper integration)
5. Language Detection: Confirms the speech is in English
6. Accent Analysis: Classifies accent based on acoustic features
7. Result Generation: Returns accent type, confidence score, and explanation

Tech Stack:

• Frontend: Streamlit (Python web framework)
• Audio Processing: FFmpeg, Librosa, PyDub
• Video Download: yt-dlp (YouTube), requests (direct URLs)
• Machine Learning: PyTorch, Transformers
• Deployment: Docker, Streamlit Cloud, Heroku

Limitations

• Simulated accent detection: Current implementation uses feature-based classification
• Speech recognition: Uses placeholder for transcription (ready for Whisper integration)
• Processing time: Large videos may take several minutes
• Audio quality dependency: Poor audio quality affects accuracy
• Single speaker optimization: Works best with one primary speaker

Future Improvements for Production

1. Enhanced Accent Detection: Replace simulation with trained accent classification model
2. Improved Speech Recognition: Integrate OpenAI Whisper or similar model
3. Real-time Processing: Add streaming audio processing capabilities
4. Batch Processing: Support multiple video analysis
5. API Integration: Create REST API for HR system integration
6. Advanced Analytics: Detailed accent characteristic breakdown
7. Database Storage: Store analysis results and user sessions

Ethical Considerations

This tool is designed as a preliminary screening aid and should be used responsibly:

• Not a sole decision factor: Use alongside human evaluation
• Bias awareness: Be conscious of potential algorithmic biases
• Privacy protection: Ensure candidate data is handled securely
• Legal compliance: Follow local hiring and discrimination laws
• Transparency: Inform candidates about automated screening tools

License

MIT License - feel free to modify and distribute.

Support

For issues or questions:

1. Check the troubleshooting section above
2. Ensure all prerequisites are properly installed
3. Test with a simple, short video first
4. Create an issue on the repository for bug reports

---