heartmula

Generate music from lyrics and tags using HeartMuLa on CUDA GPUs.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/Hermesagents/hermes-agents --skill heartmula-hermesagents
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/Hermesagents/hermes-agents/tree/main/skills/media/heartmula
Command: npx skills add https://github.com/Hermesagents/hermes-agents --skill heartmula-hermesagents

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

HeartMuLa provides an open-source pathway to generate full songs from simple lyrics and descriptive tags, reducing dependency on proprietary models and enabling offline experimentation.

Core Features & Use Cases

  • HeartMuLa family includes a music language model (HeartMuLaGen), HeartCodec for high-fidelity audio, HeartTranscriptor for lyrics transcription, and HeartCLAP for audio-text alignment.
  • Local/offline generation on CUDA-enabled GPUs with 8GB–16GB VRAM, with CPU fallback recommended for non-GPU environments.
  • Multilingual lyric support and scalable generation pipelines for offline research and custom deployment.

Quick Start

Install HeartMuLa locally and run a sample generation with the provided scripts.

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 locally?

Generate music from lyrics locally by conditioning the model with text tags using the HeartMuLa ecosystem. It enables offline music creation on CUDA-enabled GPUs, outputting high-fidelity audio via HeartCodec.

Can I generate music offline without proprietary models?

Yes, you can generate music offline without proprietary models using this open-source ecosystem. It allows full song generation from lyrics and tags on local hardware, reducing dependency on external APIs.

What GPU VRAM is required for local music generation?

Local music generation requires a CUDA-enabled GPU with 8GB to 16GB VRAM. For non-GPU environments, a CPU fallback is recommended to run the generation pipeline.

Does the music generation model support multilingual lyrics?

Yes, the music generation model supports multilingual lyrics. It uses HeartTranscriptor for lyrics transcription and HeartCLAP for audio-text alignment to process diverse languages.

How do I install and run a sample music generation script?

Install HeartMuLa locally and run a sample generation using the provided scripts. This process involves satisfying hardware requirements, installation steps, and model checkpoint handling as detailed in the documentation.

What are the limitations of offline AI music generation?

Offline AI music generation is limited by strict hardware constraints, requiring CUDA-enabled GPUs with 8-16GB VRAM for high-fidelity audio output. Non-GPU environments must rely on a slower CPU fallback.