heartmula

name	heartmula
description	HeartMuLa: Suno-like song generation from lyrics + tags.
version	1.0.0
platforms	["linux","macos","windows"]
metadata	{"ion":{"tags":["music","audio","generation","ai","heartmula","heartcodec","lyrics","songs"],"related_skills":["audiocraft"]}}

HeartMuLa - Open-Source Music Generation

Overview

HeartMuLa is a family of open-source music foundation models (Apache-2.0) designed to produce music conditioned on lyrics and tags, with multilingual capabilities. It generates complete songs from lyrics plus tags and serves as an open-source counterpart to Suno. The suite includes:

• HeartMuLa - Music language model (3B/7B) that generates audio from lyrics + tags
• HeartCodec - 12.5Hz music codec for high-fidelity audio reconstruction
• HeartTranscriptor - Whisper-based lyrics transcription
• HeartCLAP - Audio-text alignment model

When to Use

• The user wants to create music or songs from textual descriptions
• The user is looking for an open-source alternative to Suno
• The user needs local/offline music generation
• The user asks about HeartMuLa, heartlib, or AI-driven music generation

Hardware Requirements

• Minimum: 8GB VRAM with --lazy_load true (loads/unloads models sequentially)
• Recommended: 16GB+ VRAM for comfortable single-GPU usage
• Multi-GPU: Use --mula_device cuda:0 --codec_device cuda:1 to split across GPUs
• 3B model with lazy_load peaks at ~6.2GB VRAM

Installation Steps

1. Clone Repository

cd ~/  # or desired directory
git clone https://github.com/HeartMuLa/heartlib.git
cd heartlib

2. Create Virtual Environment (Python 3.10 required)

uv venv --python 3.10 .venv
. .venv/bin/activate
uv pip install -e .

3. Fix Dependency Compatibility Issues

IMPORTANT: As of Feb 2026, the pinned dependencies have conflicts with newer packages. Apply these fixes:

# Upgrade datasets (old version incompatible with current pyarrow)
uv pip install --upgrade datasets

# Upgrade transformers (needed for huggingface-hub 1.x compatibility)
uv pip install --upgrade transformers

4. Patch Source Code (Required for transformers 5.x)

Patch 1 - RoPE cache fix in src/heartlib/heartmula/modeling_heartmula.py:

Inside the setup_caches method of the HeartMuLa class, insert RoPE reinitialization after the reset_caches try/except block and before the with device: block:

# Re-initialize RoPE caches that were skipped during meta-device loading
from torchtune.models.llama3_1._position_embeddings import Llama3ScaledRoPE
for module in self.modules():
    if isinstance(module, Llama3ScaledRoPE) and not module.is_cache_built:
        module.rope_init()
        module.to(device)

Why: from_pretrained initially creates the model on a meta device; Llama3ScaledRoPE.rope_init() skips cache creation on meta tensors and never rebuilds them once weights are moved to a real device.

Patch 2 - HeartCodec loading fix in src/heartlib/pipelines/music_generation.py:

Add ignore_mismatched_sizes=True to every HeartCodec.from_pretrained() call (there are two: the eager load in __init__ and the lazy load in the codec property).

Why: VQ codebook initted buffers are shape [1] in the checkpoint but [] in the model. Same data, just scalar vs 0-d tensor. Safe to skip.

5. Download Model Checkpoints

cd heartlib  # project root
hf download --local-dir './ckpt' 'HeartMuLa/HeartMuLaGen'
hf download --local-dir './ckpt/HeartMuLa-oss-3B' 'HeartMuLa/HeartMuLa-oss-3B-happy-new-year'
hf download --local-dir './ckpt/HeartCodec-oss' 'HeartMuLa/HeartCodec-oss-20260123'

All three downloads can run in parallel. Total size is several GB.

GPU / CUDA

HeartMuLa defaults to CUDA (--mula_device cuda --codec_device cuda). No extra configuration is needed when the system has an NVIDIA GPU with PyTorch CUDA support.

• The bundled torch==2.4.1 ships with CUDA 12.1 support out of the box
• torchtune may report version 0.4.0+cpu — this is purely package metadata; it still delegates to CUDA through PyTorch
• To confirm GPU usage, look for "CUDA memory" lines in the output (e.g. "CUDA memory before unloading: 6.20 GB")
• No GPU? CPU execution is possible with --mula_device cpu --codec_device cpu, but expect generation to be extremely slow (potentially 30-60+ minutes for a single song vs ~4 minutes on GPU). CPU mode also demands significant RAM (~12GB+ free). If the user lacks an NVIDIA GPU, suggest a cloud GPU service (Google Colab free tier with T4, Lambda Labs, etc.) or the online demo at https://heartmula.github.io/ instead.

Usage

Basic Generation

cd heartlib
. .venv/bin/activate
python ./examples/run_music_generation.py \
  --model_path=./ckpt \
  --version="3B" \
  --lyrics="./assets/lyrics.txt" \
  --tags="./assets/tags.txt" \
  --save_path="./assets/output.mp3" \
  --lazy_load true

Input Formatting

Tags (comma-separated, no spaces):

piano,happy,wedding,synthesizer,romantic

or

rock,energetic,guitar,drums,male-vocal

Lyrics (use bracketed structural tags):

[Intro]

[Verse]
Your lyrics here...

[Chorus]
Chorus lyrics...

[Bridge]
Bridge lyrics...

[Outro]

Key Parameters

Parameter	Default	Description
--max_audio_length_ms	240000	Max length in ms (240s = 4 min)
--topk	50	Top-k sampling
--temperature	1.0	Sampling temperature
--cfg_scale	1.5	Classifier-free guidance scale
--lazy_load	false	Load/unload models on demand (saves VRAM)
--mula_dtype	bfloat16	Dtype for HeartMuLa (bf16 recommended)
--codec_dtype	float32	Dtype for HeartCodec (fp32 recommended for quality)

Performance

• RTF (Real-Time Factor) ≈ 1.0 — a 4-minute song takes ~4 minutes to generate
• Output: MP3, 48kHz stereo, 128kbps

Pitfalls

1. Do NOT use bf16 for HeartCodec — degrades audio quality. Use fp32 (default).
2. Tags may be ignored — known issue (#90). Lyrics tend to dominate; experiment with tag ordering.
3. Triton not available on macOS — Linux/CUDA only for GPU acceleration.
4. RTX 5080 incompatibility reported in upstream issues.
5. The dependency pin conflicts require the manual upgrades and patches described above.

Links

• Repo: https://github.com/HeartMuLa/heartlib
• Models: https://huggingface.co/HeartMuLa
• Paper: https://arxiv.org/abs/2601.10547
• License: Apache-2.0