Environmental MusicGen
LoRA fine-tuned Meta's MusicGen for ambient environmental music
Final project for a generative AI class. Implemented LoRA (Low-Rank Adaptation) and fine-tuned Meta’s MusicGen model to generate ambient environmental music hybrids.
The model learns to combine environmental sounds with musical elements, creating unique ambient compositions that blend nature sounds with melodic structures.
Tech Stack: Python, PyTorch, Meta MusicGen, LoRA, Hugging Face
Key Contributions:
- Implemented LoRA fine-tuning pipeline for MusicGen
- Curated training dataset of environmental sounds and ambient music
- Evaluated generated outputs using perceptual metrics