Speech Processing Service

A comprehensive web application for audio transcription, text-to-speech conversion, and document processing with GPU-accelerated performance.

Features

• 🎙️ Audio/Video Transcription - Support for MP3, WAV, MP4, MOV files
• 📺 YouTube Processing - Extract transcripts from videos and playlists
• 📄 Document Processing - Extract text from PDF, DOCX, and TXT files
• 🔊 Text-to-Speech - Generate audio with multiple voice options
• 🌍 Multi-Language Support - 10 languages including English, Spanish, French, German, and more
• 📊 Multiple Output Formats - Text, Markdown, Word documents, and PDF
• ⚡ GPU Acceleration - Powered by faster-whisper for efficient transcription
• 📈 Real-time Progress - Track processing status with visual indicators
• 💾 Job History - All processing jobs saved to database

Quick Start

Prerequisites

• Python 3.11+
• PostgreSQL database
• FFmpeg
• faster-whisper (for transcription)

Local Development

1. Clone the repository

   git clone https://github.com/YOUR_USERNAME/YOUR_REPO.git
   cd YOUR_REPO

2. Install dependencies

   pip install -r requirements.txt

3. Set up environment variables

   cp .env.example .env
   # Edit .env with your configuration

4. Run the application

   gunicorn --bind 0.0.0.0:5000 main:app

5. Access the app Open http://localhost:5000 in your browser

Deployment

Deploy to Ubuntu Server

For detailed deployment instructions, see deployment/DEPLOY_TO_UBUAI.md

Quick deployment on Ubuntu:

# On your Ubuntu server
sudo mkdir -p /var/www
sudo git clone https://github.com/YOUR_USERNAME/YOUR_REPO.git /var/www/speech-app
cd /var/www/speech-app
sudo bash deployment/setup_ubuntu.sh

Deployment Features

• ✅ Nginx reverse proxy
• ✅ Gunicorn WSGI server
• ✅ Systemd service management
• ✅ PostgreSQL database
• ✅ Automatic service restart
• ✅ Log rotation
• ✅ Security hardening

Usage

Audio Transcription

1. Select the Files tab
2. Upload your audio/video file (MP3, WAV, MP4, MOV)
3. Choose the language
4. Select output format
5. Click "Process"

YouTube Transcription

1. Select the YouTube tab
2. Paste a YouTube video or playlist URL
3. Choose whether to use existing transcript or transcribe audio
4. Select language and output format
5. Click "Process"

Document Processing

1. Select the Documents tab
2. Upload a PDF, DOCX, or TXT file
3. Select output format
4. Click "Process"

Text-to-Speech

1. Select the Text tab
2. Enter or paste your text
3. Choose a voice and language
4. Click "Generate Speech"

Architecture

• Backend: Flask + SQLAlchemy
• Database: PostgreSQL
• Transcription: faster-whisper (GPU-accelerated)
• Web Server: Gunicorn + Nginx
• TTS: gTTS + pyttsx3
• Video Processing: yt-dlp + moviepy

Configuration

Environment Variables

# Database
DATABASE_URL=postgresql://user:password@localhost/dbname

# Whisper Configuration
WHISPER_SERVER=localhost
WHISPER_SCRIPT_PATH=/path/to/faster-whisper/script

# Application
FLASK_ENV=production
SECRET_KEY=your-secret-key

Supported Languages

• English (en)
• Spanish (es)
• French (fr)
• German (de)
• Italian (it)
• Portuguese (pt)
• Russian (ru)
• Japanese (ja)
• Korean (ko)
• Chinese (zh)

Output Formats

• Plain Text (.txt)
• Markdown (.md)
• Word Document (.docx)
• PDF (.pdf)

API Integration

faster-whisper GPU Server

The application integrates with a GPU-accelerated faster-whisper instance:

• Primary: Remote GPU server via SSH
• Fallback: Local faster-whisper installation
• Auto-detection of best processing method

Troubleshooting

Common Issues

Service won't start:

sudo journalctl -u speech-app -f

Database connection errors:

sudo systemctl status postgresql

Transcription fails:

• Check faster-whisper installation
• Verify GPU server connectivity
• Review application logs

Logs

• Application: /var/log/speech-app/
• Nginx: /var/log/nginx/
• System: sudo journalctl -u speech-app

Development

Project Structure

.
├── app.py                 # Flask application setup
├── main.py               # Application entry point
├── models.py             # Database models
├── utils/
│   ├── audio_converter.py
│   ├── youtube_processor.py
│   ├── document_processor.py
│   ├── text_to_speech.py
│   ├── output_formatter.py
│   └── whisper_client.py
├── templates/            # HTML templates
├── static/              # CSS, JS, assets
└── deployment/          # Deployment scripts

Adding Features

1. Create utility modules in utils/
2. Add routes in app.py
3. Update models in models.py
4. Add templates in templates/

Service Management

Commands

# Check status
sudo systemctl status speech-app

# Restart service
sudo systemctl restart speech-app

# View logs
sudo journalctl -u speech-app -f

# Update application
cd /var/www/speech-app
git pull
sudo systemctl restart speech-app

Performance

• GPU-accelerated transcription with faster-whisper
• Asynchronous job processing
• Efficient file handling for large uploads
• Database query optimization
• Nginx caching for static assets

Security

• Environment-based secret management
• SQL injection protection via SQLAlchemy
• XSS protection headers
• CSRF protection
• File upload validation
• Secure password handling

License

This project is for internal use.

Support

For deployment assistance, see:

• Ubuntu Deployment Guide
• UBUAI Server Guide
• Deployment README

Credits

Built with:

• Flask
• faster-whisper
• PostgreSQL
• Nginx
• Gunicorn