Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/greydgl/musicai


https://github.com/greydgl/musicai

Last synced: 4 months ago
JSON representation

Awesome Lists containing this project

README 

          
MusicAI - AI Music Generation & Evaluation System

==================================================



A comprehensive web-based system for generating and evaluating AI-generated music using Meta's MusicGen model with structured evaluation criteria.



REPOSITORY

----------

https://github.com/GreyDGL/MusicAI



OVERVIEW

--------

MusicAI combines state-of-the-art music generation with professional evaluation tools, enabling collaborative research and assessment of AI-generated music quality.



FEATURES

--------



Music Generation:

- Real MusicGen Integration: Uses Meta's MusicGen model for high-quality music generation

- Multiple Model Sizes: Small (300MB), Medium (1.5GB), Large (3.3GB), and Melody variants

- Customizable Parameters: Duration (1-30s), Temperature, Guidance Scale

- Text-to-Music: Generate music from natural language descriptions

- CLI Support: Generate music via command line or web interface



Music Evaluation System:

- Structured Criteria: Professional evaluation framework with 6 key aspects:

  * Melodic Content (clarity, development, thematic material)

  * Instrumentation/Timbre (authenticity, clarity, synthetic elements)

  * Rhythmic Structure (groove, timing, forward motion)

  * Mood/Emotional Alignment (prompt matching, emotional impact)

  * Audio Quality (artifacts, balance, clarity)

  * Overall Rating

- Visual Waveform Display: Interactive audio visualization with WaveSurfer.js

- Star Rating System: 1-5 scale for quantitative assessment

- Detailed Comments: Free-form feedback for qualitative insights



Collaboration & Sync:

- Git-Based Syncing: Share evaluations through GitHub

- Database Synchronization: All evaluations stored in SQLite database

- Multi-User Support: Multiple evaluators can work on the same dataset

- Export Functionality: Export all data as JSON for analysis



Dashboard & Analytics:

- Overview Statistics: Total files, evaluations, average ratings

- Rating Distribution: Visual charts showing rating patterns

- Top-Rated Music: Leaderboard of best-performing generations

- Recent Activity: Track latest evaluations and generations



PREREQUISITES

-------------

- Python 3.8 or higher

- Git

- 4GB+ RAM (8GB+ recommended for Medium/Large models)

- macOS, Linux, or Windows

- ~10GB free disk space (for model downloads)



INSTALLATION

------------



Automated Installation (Recommended):

1. Clone the repository:

   git clone https://github.com/GreyDGL/MusicAI.git

   cd MusicAI



2. Run installation script:

   ./install.sh  (macOS/Linux)

   OR

   python install.py  (Windows)



3. Start the application:

   python app.py



Manual Installation:

1. Clone repository:

   git clone https://github.com/GreyDGL/MusicAI.git

   cd MusicAI



2. Create virtual environment:

   python -m venv venv

   source venv/bin/activate  (macOS/Linux)

   OR

   venv\Scripts\activate  (Windows)



3. Install dependencies:

   pip install -r requirements.txt

   pip install -r requirements_musicgen.txt



4. Start application:

   python app.py



USAGE

-----



Web Interface:

1. Start the application: python app.py

2. Open browser: http://localhost:8080

3. Navigate through tabs:

   - Generate: Create new music from text prompts

   - Evaluate: Listen and evaluate music files

   - Dashboard: View statistics and analytics

   - Export: Download evaluation data



Command Line Interface:

   # Generate music

   python musicgen_cli.py "A peaceful piano melody"



   # With options

   python musicgen_cli.py "Epic orchestral music" --duration 15 --model medium



   # List available models

   python musicgen_cli.py --list-models



Git Synchronization:

   # Get latest updates

   git pull origin main



   # Save your work

   git add .

   git commit -m "Added evaluations for jazz pieces"

   git push origin main



PROJECT STRUCTURE

-----------------

MusicAI/

├── app.py                    # Main Flask application

├── models.py                 # Database models (SQLAlchemy)

├── config.py                 # Configuration settings

├── musicgen_api.py          # MusicGen integration

├── musicgen_cli.py          # CLI for music generation

├── requirements.txt         # Python dependencies

├── requirements_musicgen.txt # MusicGen dependencies

├── install.sh               # Installation script

├── cleanup.sh               # Cleanup utility

├── run.sh                   # Launch script

├── music/                   # Generated music storage

├── instance/               # Instance-specific files

│   └── database.db        # SQLite database

├── static/                 # Frontend assets

│   ├── css/

│   │   └── style.css

│   └── js/

│       └── main.js

└── templates/              # HTML templates

    ├── base.html

    ├── index.html         # Dashboard

    ├── generate.html      # Generation interface

    └── evaluate.html      # Evaluation interface



CONFIGURATION

-------------



config.py Settings:

- PORT: Application port (default: 8080)

- MUSIC_FOLDER: Music storage location

- DATABASE_URL: Database connection

- MAX_DURATION: Maximum generation duration

- DEFAULT_MODEL: Default MusicGen model size



Model Performance:

+--------+-------+-----------+-----------------+---------+

| Model  | Size  | RAM Usage | Generation Time | Quality |

+--------+-------+-----------+-----------------+---------+

| Small  | 300MB | ~2GB      | 10-20s          | Good    |

| Medium | 1.5GB | ~6GB      | 20-40s          | Better  |

| Large  | 3.3GB | ~12GB     | 30-60s          | Best    |

+--------+-------+-----------+-----------------+---------+



EXAMPLE PROMPTS

---------------

- "Upbeat jazz piano with walking bass line"

- "Cinematic orchestral piece with strings and brass"

- "Lo-fi hip hop beat for studying"

- "Classical guitar with Spanish influences"

- "Ambient electronic with nature sounds"

- "90s rock anthem with power chords"

- "Smooth R&B with soul vocals feel"

- "Traditional Japanese music with koto and flute"



TROUBLESHOOTING

---------------



Common Issues:

- Port already in use: Change PORT in config.py

- Memory errors: Use smaller model or reduce duration

- Model download fails: Check internet connection and disk space

- Can't find Python: Use python3 instead of python

- Permission denied: Run with appropriate permissions or check file ownership



Reset & Cleanup:

   # Clean up and reset

   ./cleanup.sh



   # Remove models cache

   rm -rf ~/.cache/huggingface/hub/



COLLABORATION GUIDE

-------------------



For Project Owner:

1. Create GitHub repository

2. Push initial code

3. Add collaborators in Settings → Manage access

4. Share repository URL



For Collaborators:

1. Accept GitHub invitation

2. Clone repository

3. Make evaluations

4. Commit and push changes

5. Pull updates regularly



DEVELOPMENT

-----------



API Endpoints:

- GET / - Dashboard

- GET /generate - Generation page

- GET /evaluate - Evaluation page

- POST /api/generate - Generate music

- POST /api/evaluate - Submit evaluation

- GET /api/music/list - List music files

- GET /api/music//stream - Stream audio

- GET /api/export/evaluations - Export data



Database Schema:

- MusicFile: Stores generated music metadata

- Evaluation: Stores evaluation data and ratings



LICENSE

-------

This project uses open-source components including Meta's MusicGen model.



ACKNOWLEDGMENTS

---------------

- Meta AI for MusicGen model

- Flask community for web framework

- Bootstrap for UI components

- WaveSurfer.js for audio visualization



SUPPORT

-------

For issues or questions, please open an issue at: https://github.com/GreyDGL/MusicAI/issues