heartmula

Generate full songs from lyrics and tags using HeartMuLa on local hardware.

Updated Mar 24, 2026
One-click install
npx skills add https://github.com/1thirteeng3/greenmoire --skill heartmula-1thirteeng3
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/1thirteeng3/greenmoire/tree/main/integrations/hermes-agent/skills/media/heartmula
Command: npx skills add https://github.com/1thirteeng3/greenmoire --skill heartmula-1thirteeng3

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

HeartMuLa provides an open-source foundation for generating full songs from lyrics and descriptive tags, enabling offline music creation without relying on proprietary services.

Core Features & Use Cases

  • HeartMuLa: music language model (3B/7B) for generation from lyrics and tags
  • HeartCodec: a 12.5Hz codec for high-fidelity audio reconstruction
  • HeartTranscriptor / HeartCLAP: lyrics transcription and audio-text alignment
  • Use Case: create multilingual songs by feeding lyrics and tags and generating full MP3 outputs on local hardware.

Quick Start

Run the example generator to produce a track from your lyrics and tags using HeartMuLa on local hardware.

Frequently Asked Questions about heartmula

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

FAQPage Schema
How do I generate music from lyrics and tags offline?

To generate music from lyrics and tags offline, you run the HeartMuLa model family on local hardware to compose multilingual songs into full MP3 outputs. You provide the lyrics and descriptive tags, and the model generates the audio track.

What hardware do I need for local music generation?

Local music generation requires Python 3.10+ and CUDA-capable GPUs, though a CPU fallback is supported. You also need to download the supplied model checkpoints and configuration files to run the generation process on your machine.

Can I generate multilingual songs using an open-source model?

Yes, you can generate multilingual songs using the open-source HeartMuLa model. It is designed to take your lyrics and descriptive tags as input to synthesize full song audio without relying on proprietary or cloud services.

Does offline music generation support multi-device setups?

Offline music generation supports optional multi-device setups, lazy loading, and patch steps as described in the configuration. These options allow you to optimize how the HeartMuLa model family utilizes your local hardware resources.

What components are used for high-fidelity audio reconstruction?

High-fidelity audio reconstruction uses HeartCodec, a 12.5Hz codec, alongside HeartTranscriptor and HeartCLAP for lyrics transcription and audio-text alignment. These components work together to ensure the generated audio matches the input lyrics.