license: mit
Model Overview
The Melody Guided Music Generation (MG²) model is an innovative approach that uses melody to guide music generation, achieving impressive results despite its simplicity and minimal resource requirements. MG² aligns melody with audio waveforms and text descriptions via a multimodal alignment module and conditions its diffusion module on these learned melody representations. This enables MG² to create music that matches the style of given audio and reflects the content of text descriptions.
Demo
Explore the capabilities of the MG² model through an online demo:
- Demo Link: Model Demo
- Instructions: Input a text description, then click "Generate" to see the music generated by the model.
GitHub Repository
Access the code and additional resources for the MG² model:
- GitHub Link: MG² on GitHub
Integration with Transformers and Hugging Face Hub
We are currently working on integrating MG² into the Hugging Face Transformers library and making it available on the Hugging Face Hub 🤗.
Tips: To generate high-quality music using MG², you'd better craft detailed and descriptive prompts that provide rich context and specific musical elements.
Paper
Title: "Melody Is All You Need For Music Generation" Authors: Shaopeng Wei, Manzhen Wei, Haoyu Wang, Yu Zhao, Gang Kou Year: 2024 arXiv Link
Citation
@article{wei2024melodyneedmusicgeneration,
title={Melody Is All You Need For Music Generation},
author={Shaopeng Wei and Manzhen Wei and Haoyu Wang and Yu Zhao and Gang Kou},
year={2024},
eprint={2409.20196},
archivePrefix={arXiv},
primaryClass={cs.SD},
url={https://arxiv.org/abs/2409.20196},
}