Hugging Peers

MiniMax-Music3-nvfp4

☆ Save
by mlx-community text-to-audio mlx-audio other
90 downloads 2 likes on Hugging Face updated

Model card

MiniMax Music 3 · MLX NVFP4

Native MLX NVFP4 weights for MiniMaxAI/MiniMax-Music3, converted for lyric-conditioned song generation on Apple Silicon with mlx-audio. This variant is experimental until it receives broader listening evaluation.

Community conversion, not an official MiniMax release. All model credit goes to MiniMax. Review the original model card and license before use.

Other MLX variants: BF16 · 8-bit · 6-bit · 4-bit · MXFP8 · MXFP4 (experimental)

Install

MiniMax Music 3 support was merged upstream in Blaizzy/mlx-audio#888. Until a PyPI release includes it, install the upstream merge commit directly:

python -m pip install "mlx-audio @ git+https://github.com/Blaizzy/mlx-audio.git@784b29e2691a93ca7483147d86f61859dfaa6296"

Generate

python -m mlx_audio.music.generate \
  --model mlx-community/MiniMax-Music3-nvfp4 \
  --caption "Warm acoustic pop, 96 BPM, intimate female vocal" \
  --lyrics $'[verse]\nMorning light across the room\n[chorus]\nSing with me' \
  --duration 30 \
  --steps 30 \
  --seed 7 \
  --output song.wav



from mlx_audio.music import load

model = load("mlx-community/MiniMax-Music3-nvfp4")
result = next(
    model.generate(
        text="Warm acoustic pop, 96 BPM, intimate female vocal",
        lyrics="[verse]\nMorning light across the room\n[chorus]\nSing with me",
        duration=30,
        steps=30,
        seed=7,
    )
)
print(result.audio.shape, result.sample_rate)  # stereo, 44100 Hz

Lyrics are required by the checkpoint contract. Use [instrumental] explicitly for instrumental generation. Duration is a requested upper bound: the autoregressive stage may emit its end token early. Style, tempo, instrument, and vocal controls are probabilistic rather than strict.

Quantization and verification

  • NVFP4 weight quantization with group size 16.
  • Large linears in the global language model, RVQ depth decoder, and flow transformer are quantized. Embeddings, output heads, convolutions, condition encoder, and vocoder remain dense for fidelity.
  • The two weight shards total 9,199,077,544 bytes.
  • Strict loading and real generation produced finite 44.1 kHz stereo audio.
  • The mlx-audio regression suite passed 1,742 tests with 34 expected skips; the focused music, converter, and registry suite passed 43 tests and 3 subtests.

Converted with mlx-audio 0.4.8 development commit c2fa486 and MLX 0.31.2.

License

The weights remain subject to the MiniMax-Music3 Community License, including its acceptable-use and commercial terms. The full license text is included in this repository.