HeartMuLa
Generation and understanding, featuring high-fidelity song synthesis and controllable structural creation
Note: Please open in the PC Client. Go to download >>>
Features
Supported Languages
System Requirements
32GB RAM recommended. 30GB+ storage recommended.
macOS 15+: M-series chips required.
Windows 10/11: NVIDIA GPU (12GB+ VRAM) recommended. Intel or AMD GPU compatibility unverified.
Note: For NVIDIA GPUs, install a newer driver.Screenshots
Introduction
HeartMuLa is a versatile "AI Music Virtuoso" that understands and creates music across various cultural boundaries:
- Multilingual Expertise: Unlike many tools, it features robust multilingual support, including but not limited to English, Chinese, Japanese, Korean, and Spanish.
- Text-to-Song: Simply provide lyrics or descriptions, and it generates high-quality songs with synthesized vocals and full instrumentation.
- Structural Control: You can act as a director, specifying the musical energy and style for different sections (e.g., Verse, Chorus, Outro).
- Lyric Transcription: It can "listen" to complex audio tracks and accurately extract lyrics across different languages.
Key Features & Capabilities
- Global Language Support: Seamlessly handles prompts and lyrics in English, Chinese, Japanese, Korean, Spanish, and more.
- All-in-One Framework: Integrates music generation, understanding, lyric recognition, and audio-text alignment into a single library.
- Pro-Level Quality: Aims to match leading commercial AI services (like Suno) in terms of acoustic fidelity and musicality.
- Open Source: The code and model weights are released to the community, fostering transparency and local innovation.
Team & Core Technology
The Team: HeartMuLa is the result of a collaborative effort between leading academic and research institutions, including:
Peking University
The Chinese University of Hong Kong
Scale Global / Ario
Contributions also include expertise from Independent Researchers.
Underlying Technology:
HeartMuLa LLM: A large language model architecture that treats music generation as a sophisticated sequence modeling task.
HeartCodec: A proprietary high-fidelity audio codec that ensures crystal-clear sound output.
HeartCLAP: A cross-modal alignment technology that bridges the gap between human language and musical audio.