What problem does it solve? Setting up and running the HeartMuLa open-source music generation model family involves dependency conflicts, source code patches, and multi-gigabyte model downloads that are easy to get wrong without guidance. ## Core Features & Use Cases - Guided Installation: Walks through cloning heartlib, creating a Python 3.10 environment, fixing dependency conflicts, and applying required source patches for transformers 5.x compatibility. - Song Generation: Generates complete songs (MP3, 48kHz stereo) from bracketed lyrics files and comma-separated style tags using the 3B/7B HeartMuLa models with HeartCodec audio reconstruction. - Hardware Planning: Covers VRAM requirements, lazy loading for 8GB GPUs, multi-GPU splitting, and CPU fallback expectations. - Use Case: A user wants an open-source Suno alternative to generate a 4-minute rock song locally from their own lyrics on a single NVIDIA GPU. ## Quick Start Set up HeartMuLa and generate a song from my lyrics.txt and tags.txt files using the 3B model with lazy loading enabled.
