agoddrie/srtgen repository overview

⁠SRTGEN - AI-Powered Subtitle Generator

⁠✨ Features

• 🎯 AI-Powered Transcription - OpenAI Whisper with GPU acceleration
• 🌍 Multi-Language Translation - Facebook's NLLB-200-1.3B for 200+ languages
• 🎬 Triple SRT Output - Original language, English, and target language
• 📝 Word-Level Timestamps - Precise subtitle timing synchronization
• 🌐 Modern Web UI - Browse files, monitor jobs, configure settings
• 🚀 Background Processing - Queue multiple transcription jobs
• 🔍 Smart File Browser - Search, filter, path memory, existing SRT detection
• 🐳 Docker Ready - One-click deployment on Unraid
• 💾 Persistent Settings - Language preferences saved in browser
• ⚡ GPU Optimized - CUDA acceleration for 10x faster processing

⁠🚀 Quick Start

⁠Docker Run (Fastest)

docker run -d --name srtgen \
  --gpus all \
  -p 5000:5000 \
  -v /path/to/media:/media \
  -e WHISPER_MODEL=medium \
  agoddrie/srtgen:latest

Access: http://localhost:5000

⁠Docker Compose (Recommended)

Create docker-compose.yml:

version: '3.8'
services:
  srtgen:
    image: agoddrie/srtgen:latest
    container_name: srtgen
    ports:
      - "5000:5000"
    volumes:
      - /mnt/user/TV-Series:/media/tv:rw
      - /mnt/user/Movies:/media/movies:rw
    environment:
      - WHISPER_MODEL=medium
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

Start:

docker-compose up -d

⁠First Run Setup

Install translation dependencies (one-time):

docker exec -it srtgen pip install transformers sentencepiece protobuf sacremoses

⁠Unraid Template

1. Go to Docker tab → Add Container
2. Click Template repositories → Add: https://github.com/agoddrie/srtgen
3. Search for SRTGEN
4. Configure paths and click Apply

⁠📋 Requirements

• NVIDIA GPU (recommended for speed)
• Docker with NVIDIA Container Runtime
• 8GB+ VRAM for medium Whisper model + NLLB-1.3B

⁠🎮 Usage

⁠Web Interface (Recommended)

1. Open browser: http://localhost:5000
2. Browse media folders or use search
3. Select video file (MKV, MP4, AVI)
4. Choose language or enable auto-detect
5. Configure overwrite settings
6. Click "Generate Subtitles"
7. Monitor progress in sidebar

⁠Output Files

For movie.mkv, generates:

movie.en.srt      # Original language transcription
movie.en.srt      # English translation (if not English)
movie.nl.srt      # Target language (e.g., Dutch)

⁠Supported Formats

• Video: .mkv, .mp4, .avi
• Languages: 99+ via Whisper (EN, NL, FR, DE, ES, IT, PT, PL, RU, JA, ZH, KO, AR, TR, etc.)
• Translation: 200+ via NLLB-200

⁠⚙️ Configuration

⁠Whisper Model Selection

Set via environment variable:

Example:

environment:
  - WHISPER_MODEL=medium

⁠Settings Modal

Access via ⚙️ icon in web UI:

• Default Language - Preset translation target
• Settings persist in browser localStorage

⁠Volume Mapping

volumes:
  - /host/path:/media/name:rw  # Read-write required for SRT creation

Important: Use :rw flag to allow subtitle file creation

⁠🛠️ Technology Stack

⁠How It Works

graph LR
    A[Video File] -->|FFmpeg| B[Audio 16kHz WAV]
    B -->|Whisper| C[Original SRT + Language]
    C -->|Whisper Translate| D[English SRT]
    C -->|NLLB-200| E[Target Language SRT]
    D --> F[Triple SRT Output]
    E --> F

1. Audio Extraction: FFmpeg → 16kHz mono WAV in /tmp
2. Transcription: Whisper with word-level timestamps + auto language detection
3. English Translation: Whisper's built-in translation task
4. Target Translation: NLLB batch processing (5 segments) with context
5. Timestamp Adjustment: Extends duration for longer translations
6. SRT Generation: Filters non-speech, formats timestamps, saves files

⁠🐛 Troubleshooting

docker exec -it srtgen pip install transformers sentencepiece protobuf sacremoses

docker run --rm --gpus all nvidia/cuda:11.8.0-base-ubuntu22.04 nvidia-smi

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

volumes:
  - /path/to/media:/media:rw  # Not :ro

⁠📦 Building from Source

git clone https://github.com/agoddrie/srtgen.git
cd srtgen
docker build -t agoddrie/srtgen:latest .
docker-compose up -d

⁠🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing)
5. Open Pull Request

⁠Development Setup

# Clone repo
git clone https://github.com/agoddrie/srtgen.git
cd srtgen

# Build image
docker build -t srtgen:dev .

# Run with hot-reload
docker run -it --rm \
  --gpus all \
  -p 5000:5000 \
  -v $(pwd)/app.py:/app/app.py \
  -v $(pwd)/mkv_transcribe.py:/app/mkv_transcribe.py \
  -v /media:/media:rw \
  srtgen:dev

⁠📊 Benchmarks

Test file: 22-minute episode (The IT Crowd S01E01)

Translation adds ~2-3 minutes for NLLB-1.3B processing

⁠📄 License

MIT License - See LICENSE⁠ file

Third-party components:

• OpenAI Whisper - MIT License⁠
• Facebook NLLB - CC BY-NC 4.0⁠
• FFmpeg - GPL/LGPL⁠

⁠🙏 Acknowledgments

• OpenAI Whisper⁠ - Robust speech recognition
• Facebook NLLB⁠ - High-quality translation
• FFmpeg⁠ - Multimedia framework
• Flask⁠ - Web framework
• PyTorch⁠ - Deep learning platform

⁠⭐ Support

If you find SRTGEN useful, please:

• ⭐ Star the repository
• 🐛 Report bugs via Issues⁠
• 💡 Request features
• 📖 Improve documentation
• 🔄 Share with others