heartmula

Generate MP3 songs from lyrics and tags using local HeartMuLa models.

174|23|Updated Apr 3, 2026
One-click install
npx skills add https://github.com/RedWoodOG/Hermes-Desktop --skill heartmula-redwoodog
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/RedWoodOG/Hermes-Desktop/tree/main/skills/media/heartmula
Command: npx skills add https://github.com/RedWoodOG/Hermes-Desktop --skill heartmula-redwoodog

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

HeartMuLa enables open-source, offline music generation by transforming lyrics and tags into full songs using local models.

Core Features & Use Cases

  • HeartMuLa model family for generation from lyrics and tags
  • HeartCodec for high-fidelity audio reconstruction
  • HeartTranscriptor for lyrics transcription
  • HeartCLAP for audio-text alignment
  • Local/offline operation with GPU acceleration
  • Supports multi-GPU setups and multilingual generation
  • Use cases include creating original songs from textual prompts and preserving privacy

Quick Start

Run the music generation script with your lyrics and tags to generate an MP3 output.

Frequently Asked Questions about heartmula

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

FAQPage Schema
How do I generate music locally from lyrics without an internet connection?

Offline music generation from lyrics uses open-source models like HeartMuLa to transform text and tags into full songs without an internet connection. It requires a Python environment and CUDA-enabled hardware to run the generation scripts and output an MP3 file.

What hardware do I need for offline AI music generation with GPU acceleration?

AI music generation with GPU acceleration requires CUDA-enabled hardware to run the models efficiently. You also need a configured Python environment and the model checkpoints to execute the generation scripts locally.

Can I generate songs in multiple languages using local open-source models?

Yes, generating songs in multiple languages using local open-source models is supported. The system handles multilingual generation, allowing you to create original songs from textual prompts in various languages while operating completely offline.

What is the best way to transcribe lyrics and align them with generated audio?

The best way to transcribe lyrics and align them with generated audio is by using dedicated components like HeartTranscriptor for transcription and HeartCLAP for audio-text alignment. These tools ensure high-fidelity audio reconstruction during the generation process.

How do I set up a multi-GPU environment for local music generation?

Setting up a multi-GPU environment for local music generation involves configuring your Python environment to support multi-GPU setups with CUDA-enabled hardware. You must install the necessary model checkpoints and dependencies as outlined in the installation guide to run HeartMuLa.

Why does offline music generation require specific Python environment setups?

Offline music generation requires a specific Python environment to manage dependencies for open-source models like HeartMuLa and HeartCodec. This setup ensures the local models can properly utilize CUDA-enabled GPU acceleration for high-fidelity audio reconstruction.