Modelpublic

mlx-community/ACE-Step1.5-MLX

sourceHugging Faceupdated 5mo agoView on Hugging Face

2likes234downloads

Model Card

ACE-Step 1.5 MLX

MLX-converted weights for ACE-Step/ACE-Step1.5, a 3.5B parameter flow-matching music generation model.

Converted using mlx-audio. Also available in 4-bit quantized.

Usage

python

from mlx_audio.tts import load

model = load("mlx-community/ACE-Step1.5-MLX")

for result in model.generate(
    text="upbeat electronic dance music with energetic synthesizers",
    duration=30.0,
):
    audio = result.audio  # [samples, 2] stereo audio
    sample_rate = result.sample_rate  # 48000

Note: The turbo model uses a 5Hz Language Model planner by default (use_lm=True) which generates a song blueprint (audio codes) before running the diffusion transformer. This is required for the turbo model to produce music.

With Vocals

python

for result in model.generate(
    text="upbeat pop song with female vocals",
    lyrics="""[verse]
Dance with me tonight
Under the neon lights

[chorus]
Move your body, feel the groove
""",
    duration=60.0,
    vocal_language="en",
):
    ...

See the ACE-Step README for full parameter documentation.