741,000+ skills indexed

xu1713COMMUNITY

Explore all skills by this author →

Verified:GitHub

heartmula

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

Updated Jun 5, 2026

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

One-click install

npx skills add https://github.com/xu1713/openhorse --skill heartmula-xu1713

Or copy as Structured Prompt for Agent▼

Please help me install this Agent Skill.
Skill: heartmula
Source: https://github.com/xu1713/openhorse/tree/main/openhorse/openhorse/skills/media/heartmula
Command: npx skills add https://github.com/xu1713/openhorse --skill heartmula-xu1713

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 provides a working local pipeline for the open-source HeartMuLa music generation models, including the dependency fixes and source patches needed to make it run. ## Core Features & Use Cases - Lyrics-to-Song Generation: Produces full MP3 songs (48kHz stereo) from bracketed lyrics and comma-separated style tags using the HeartMuLa 3B/7B language model and HeartCodec audio codec. - VRAM-Constrained Operation: Supports lazy loading to run on GPUs with as little as 8GB VRAM, plus multi-GPU splitting across devices. - Use Case: A musician wants a demo track of a romantic piano ballad. Provide lyrics with [Verse]/[Chorus] structure and tags like "piano,happy,wedding" to generate a 4-minute song locally in about 4 minutes on GPU. ## Quick Start Clone the heartlib repository, install dependencies with the documented patches, download the HeartMuLa checkpoints, and ask the AI to generate a song from your lyrics file and style tags.

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 with AI locally?▼

Use HeartMuLa by cloning the heartlib repository, installing it in a Python 3.10 virtual environment, downloading the model checkpoints, and running run_music_generation.py with a lyrics file and tags file. Output is a 48kHz stereo MP3.

What open-source alternative to Suno exists for song generation?▼

HeartMuLa is an Apache-2.0 open-source music foundation model family comparable to Suno. It generates full songs conditioned on lyrics and style tags, with 3B and 7B model variants available on Hugging Face.

How much VRAM does HeartMuLa need to run?▼

HeartMuLa requires a minimum of 8GB VRAM with lazy loading enabled, which peaks around 6.2GB for the 3B model. 16GB or more is recommended, and multi-GPU setups can split the language model and codec across devices.

Can HeartMuLa run on CPU or macOS?▼

CPU mode works via --mula_device cpu but is extremely slow, taking 30-60+ minutes per song and requiring 12GB+ RAM. Triton is unavailable on macOS, so GPU acceleration is Linux/CUDA only; consider a cloud GPU or the online demo instead.

Why does HeartMuLa fail after installing with default dependencies?▼

The pinned dependencies conflict with newer pyarrow and huggingface-hub versions, requiring upgrades to datasets and transformers. Two source patches are also needed: RoPE cache reinitialization and ignore_mismatched_sizes for HeartCodec loading.

What are the limitations of HeartMuLa song generation?▼

Style tags may be ignored since lyrics tend to dominate conditioning, a known upstream issue. RTX 5080 incompatibility has been reported, and using bf16 for the HeartCodec degrades audio quality, so fp32 is required.

Related Skills

Explore More Design & Creative Skills→

heartmula

Generate MP3 music from lyrics and tags using HeartMuLa models.

sound-design

Design layered audio plans with loudness targets and dialogue cleanup.

tianpuyue_music

Generate Tianpuyue music, songs, or lyrics and download them locally with status polling.

ai-music-generation

Generates music and songs with AudioCraft and HeartMuLa, and analyzes audio via spectrograms.