Loading...
Discovering amazing AI tools

Open-source music foundation models and generator that create full songs (melody, vocals, and lyrics) from text prompts and tags.
Open-source music foundation models and generator that create full songs (melody, vocals, and lyrics) from text prompts and tags.
HeartMuLa is a family of open-source music foundation models and tooling that generate complete songs from text or lyrics, producing melody, instrumental arrangement, and synthesized vocals. The system separates a transformer-based generation backbone (HeartMuLa) from an audio codec (HeartCodec) and includes a transcription model (HeartTranscriptor) for lyrics extraction. It supports prompt conditioning via lyrics and tags, section-level control, and multiple model variants (including 3B and RL-tuned versions) for different fidelity and performance trade-offs. HeartMuLa is designed for local inference and integration with UI projects (Gradio, ComfyUI, Next.js studios), and can download model weights from HuggingFace/ModelScope for offline use.
