754,000+ skills indexed

ShineBreakerCOMMUNITY

Explore all skills by this author →

Verified:GitHub

heartmula

Generates full songs from lyrics and style tags using HeartMuLa open-source music models.

2|Updated Jan 2, 2026

Search Tags:#music generation #open-source #heartmula #ai audio #gpu inference #song synthesis #lyrics to music

One-click install

npx skills add https://github.com/ShineBreaker/Guix-configs --skill heartmula-shinebreaker

Or copy as Structured Prompt for Agent▼

Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/ShineBreaker/Guix-configs/tree/main/dotfiles/mutable/agents/hermes/.local/share/hermes/skills/.archive/heartmula
Command: npx skills add https://github.com/ShineBreaker/Guix-configs --skill heartmula-shinebreaker

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Generating complete songs with vocals from text requires either paid services like Suno or complex local model setups. This Skill guides the installation, patching, and execution of HeartMuLa, an open-source music generation model family, so you can produce songs locally from lyrics and tags. ## Core Features & Use Cases - Song Generation from Lyrics + Tags: Produces full MP3 songs (48kHz stereo) conditioned on bracketed lyrics and comma-separated style tags using 3B/7B models. - Complete Setup Guidance: Covers cloning heartlib, Python 3.10 environment creation, dependency conflict fixes, and required source patches for transformers 5.x compatibility. - Hardware Flexibility: Supports low-VRAM mode via lazy loading (~6.2GB peak), multi-GPU splitting, and CPU fallback. - Use Case: A musician wants an open-source Suno alternative to prototype song ideas locally; this Skill walks through generating a 4-minute track from a lyrics file and tags like "piano,happy,wedding". ## Quick Start Ask the assistant to install HeartMuLa and generate a song from your lyrics file and style tags using the 3B model with lazy loading enabled.

Frequently Asked Questions about heartmula

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

FAQPage Schema

How do I generate a song from lyrics with HeartMuLa?▼

Run examples/run_music_generation.py with --model_path pointing to downloaded checkpoints, --lyrics and --tags files, and --save_path for the MP3 output. Lyrics use bracketed section tags like [Verse] and [Chorus]; tags are comma-separated without spaces.

What is a good open-source Suno alternative for local music generation?▼

HeartMuLa is an Apache-2.0 music foundation model family that generates full songs from lyrics and tags, comparable to Suno. It includes the HeartCodec audio codec, HeartTranscriptor for lyrics transcription, and HeartCLAP for audio-text alignment.

How much VRAM does HeartMuLa need to run?▼

The 3B model runs in about 6.2GB VRAM with --lazy_load true, which loads and unloads models sequentially. 16GB+ is recommended for comfortable use, and multi-GPU setups can split models via --mula_device and --codec_device.

Can HeartMuLa run on CPU or without an NVIDIA GPU?▼

Yes, use --mula_device cpu --codec_device cpu, but generation is extremely slow (30-60+ minutes per song versus ~4 minutes on GPU) and needs 12GB+ RAM. Without a GPU, consider a cloud GPU service or the online demo instead.

Why does HeartMuLa fail after installing with newer transformers?▼

The pinned dependencies conflict with newer packages, requiring upgrades to datasets and transformers plus two source patches: RoPE cache reinitialization in modeling_heartmula.py and ignore_mismatched_sizes=True in HeartCodec loading calls.

What are the limitations of HeartMuLa music generation?▼

Tags may be ignored since lyrics tend to dominate conditioning, Triton GPU acceleration is unavailable on macOS, RTX 5080 incompatibility has been reported, and bf16 for HeartCodec degrades audio quality so fp32 is required.

Related Skills

Explore More Design & Creative Skills→

minimax-music-gen

Generate vocal songs, instrumentals, and covers via the MiniMax Music API using the mmx CLI.

video-audio-extractor

Extract audio streams from video files as WAV or MP3.

acestep-lyrics-transcription

Transcribe audio files into timestamped lyrics using Whisper or ElevenLabs Scribe APIs.

heartmula

Generate music and songs from lyrics and tags using HeartMuLa models.