This repository provides tools and models for generating folk music using Long Short-Term Memory (LSTM) recurrent neural networks. It targets musicians, researchers, and enthusiasts interested in AI-driven music composition and style modeling, offering a way to create novel folk tunes and explore the intersection of traditional music and artificial intelligence.

How It Works

The project utilizes LSTMs, a type of recurrent neural network well-suited for sequential data like music. It trains on large datasets of folk music transcriptions to learn stylistic patterns, melodies, and structures. The trained models can then generate new musical pieces in a similar style, offering a deep learning approach to algorithmic composition.

Quick Start & Requirements

• Installation: Requires Python 2.7 and uses conda for environment management. Key dependencies include mkl-service, Theano (master branch), and Lasagne (master branch).
• Prerequisites: Python 2.7, conda, pip.
• Usage:
  • Generate music: python sample_rnn.py --terminal metadata/folkrnn_v2.pkl
  • Train a model: python train_rnn.py config5 data/data_v2
• Resources: Links to extensive documentation, research papers, and numerous musical examples are provided.

Highlighted Details

• Extensive academic and artistic output, including albums, performances, and research papers, demonstrating the practical application and impact of the project.
• Support for multiple versions (v1, v2, v3) of the folk-RNN models.
• Integration examples with other AI music projects like DeepBach.
• A large corpus of over 47,000 tunes available for exploration.

Maintenance & Community

The project is associated with research from institutions like Queen Mary University of London. While specific active community channels like Discord/Slack are not explicitly mentioned, the extensive list of publications and media coverage indicates significant prior engagement and research activity.

Licensing & Compatibility

The repository does not explicitly state a license. Given the nature of the dependencies (Theano, Lasagne) and the project's academic origins, users should verify licensing for commercial use or integration into closed-source projects.

Limitations & Caveats

The project relies on Python 2.7 and older deep learning libraries (Theano, Lasagne), which are largely deprecated and may present significant challenges for setup and compatibility with modern Python environments and hardware.