heartmula

Generate complete songs from lyrics and tags using HeartMuLa components.

1|Updated Apr 14, 2026
One-click install
npx skills add https://github.com/eddielueng/hermes-agent-zh --skill heartmula-eddielueng
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/eddielueng/hermes-agent-zh/tree/main/skills/media/heartmula
Command: npx skills add https://github.com/eddielueng/hermes-agent-zh --skill heartmula-eddielueng

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) components.

What problem does it solve?

HeartMuLa removes the hassle of building and configuring an open-source, local music generation pipeline by providing end-to-end steps to set up HeartMuLa and run song generation from text.

Core Features & Use Cases

  • Song generation from lyrics + tags: Produces complete songs conditioned on multilingual lyrics and comma-separated tags.
  • End-to-end open-source stack: Works with HeartMuLa (generation), HeartCodec (audio reconstruction), HeartTranscriptor (Whisper-based transcription), and HeartCLAP (audio-text alignment).
  • Local/offline workflow: Enables on-device music creation comparable to commercial “Suno-like” experiences without SaaS lock-in.
  • Practical hardware scaling: Supports single-GPU and multi-GPU configurations, with lazy loading to reduce VRAM pressure.

Quick Start

Run heartmula by generating from your lyrics and tags using the HeartMuLa checkout and checkpoints (set model_path to your downloaded HeartMuLa ckpt files, point lyrics/tags to your text files, and save the result mp3).

Frequently Asked Questions about heartmula

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a full song from lyrics and tags locally?

You can generate a full song by providing multilingual lyrics and comma-separated tags to the pipeline. It uses open-source components to condition the generation and outputs a complete MP3 file.

Can I run offline text-to-music generation without SaaS lock-in?

Yes, you can run offline text-to-music generation on-device without SaaS lock-in. This approach provides a Suno-like experience using local checkpoints and an end-to-end open-source stack for audio creation.

What do I need to set up for local music generation with HeartMuLa?

Local music generation requires installing the heartlib environment, downloading HeartMuLa and HeartCodec checkpoints, and applying documented dependency and transformer patches to ensure a working end-to-end pipeline.

Does this open-source audio generation pipeline support multi-GPU configurations?

Yes, this open-source audio generation pipeline supports both single-GPU and multi-GPU configurations. It also includes optional lazy loading to help reduce VRAM pressure during the song generation process.

How does the audio reconstruction process work in this music generation stack?

The audio reconstruction process works by using HeartCodec to rebuild audio from generated outputs. It is integrated with HeartTranscriptor for Whisper-based transcription and HeartCLAP for audio-text alignment.