musicgen-unified 2.0.0

Simple, clean instrumental music generation with MusicGen

MusicGen Unified - v2.0

A clean, focused implementation of Facebook's MusicGen for instrumental music generation.

What This Is

• Simple: Just music generation, no unnecessary complexity
• Fast: GPU-optimized for 10x realtime generation
• Practical: CLI, API, and web interface included
• Deployable: One-click AWS deployment

What This Isn't

• Not for vocal/singing generation (MusicGen doesn't support it)
• Not a microservices architecture (it's a simple tool)
• Not a research framework (it's for practical use)

Quick Start

# Install
pip install musicgen-unified

# Generate music
musicgen generate "upbeat jazz piano" --duration 30

# Start web interface
musicgen serve

# Start API
musicgen api

Features

• ✅ Text-to-music generation
• ✅ Extended generation (>30 seconds)
• ✅ Batch processing from CSV
• ✅ GPU acceleration
• ✅ Web interface
• ✅ REST API
• ✅ AWS deployment ready

Installation

# Basic installation
pip install musicgen-unified

# With GPU support
pip install musicgen-unified[gpu]

# Development
git clone https://github.com/yourusername/musicgen-unified
cd musicgen-unified
pip install -e ".[dev]"

Usage

Command Line

# Basic generation
musicgen generate "smooth jazz saxophone" -o jazz.mp3

# Extended generation
musicgen generate "epic orchestral" --duration 120 -o epic.mp3

# Batch processing
musicgen batch playlist.csv --output-dir music/

Python API

from musicgen import MusicGenerator

# Initialize
generator = MusicGenerator(device="cuda")

# Generate music
audio, sample_rate = generator.generate(
    "ambient electronic",
    duration=30.0
)

# Save
generator.save_audio(audio, sample_rate, "output.mp3")

Web Interface

musicgen serve --port 8080
# Open http://localhost:8080

REST API

musicgen api --port 8000

# Generate music
curl -X POST http://localhost:8000/generate \
  -H "Content-Type: application/json" \
  -d '{"prompt": "classical piano", "duration": 30}'

Deployment

Docker

docker build -t musicgen .
docker run -p 8080:8080 --gpus all musicgen

AWS

# Configure AWS credentials
aws configure

# Deploy
./deployment/deploy_aws.sh

Architecture

musicgen-unified/
├── musicgen/
│   ├── generator.py    # Core generation with GPU optimization
│   ├── batch.py        # Batch processing
│   ├── prompt.py       # Prompt engineering
│   ├── api.py          # FastAPI server
│   ├── cli.py          # CLI interface
│   └── utils.py        # Utilities
├── static/             # Web UI files
├── tests/              # Test suite
└── deployment/         # Deployment scripts

Performance

• GPU: 10x faster than realtime
• CPU: 1x realtime (approximate)
• Memory: <1GB base, ~4GB with large model
• Startup: <10 seconds

Models

• small: Fast, lower quality (300M parameters)
• medium: Balanced (1.5B parameters)
• large: Best quality (3.3B parameters)

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE file for details.

Acknowledgments

• Facebook Research for MusicGen
• The open source community

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: [e-mail]