You can deploy this model and start earning money today!

Abiray/
MiniMax-Music3-GGUF

Name: MiniMax-Music3-GGUF
Availability: InStock

MiniMaxAI/MiniMax-Music3

quantized

apache-2.0

Views

No views yet

MiniMax-Music-3 GGUF Quantizations

This repository contains GGUF quantizations for the diffusion transformer / DiT backbone of MiniMax-Music3, a state-of-the-art text-to-music model capable of generating structured, full-length songs up to 5 minutes long at 32 kHz stereo.

These quantized weights enable running MiniMax-Music-3 locally inside ComfyUI using the ComfyUI-GGUF custom node with significantly lower VRAM requirements.

📦 Quantized Files & Recommendations

File Name	Quantization	Size	Recommended Use
`MiniMax-Music3-F16.gguf`	F16	4.98 GB	Uncompressed reference baseline
`MiniMax-Music3-Q8_0.gguf`	Q8_0	2.70 GB	Near-lossless audio fidelity (Recommended)
`MiniMax-Music3-Q6_K.gguf`	Q6_K	2.12 GB	Excellent fidelity, minimal compression loss
`MiniMax-Music3-Q5_K_M.gguf`	Q5_K_M	1.79 GB	Great balance of speed and VRAM footprint
`MiniMax-Music3-Q4_K_M.gguf`	Q4_K_M	1.49 GB	Low VRAM setups (8GB–10GB GPUs)
`MiniMax-Music3-Q3_K_M.gguf`	Q3_K_M	1.16 GB	Ultra-low VRAM (May introduce audible softening)

🧩 Required Companion Files

The GGUF files in this repository replace only the diffusion transformer (DiT). To generate music in ComfyUI, you will also need the text encoder and VAE models provided by Comfy-Org:

Text Encoders: Comfy-Org/MiniMax-Music-3 Text Encoders
VAE: Comfy-Org/MiniMax-Music-3 VAE

📂 ComfyUI File Placement

Place the downloaded files into their respective folders in your ComfyUI installation:

text

ComfyUI/
└── models/
    ├── diffusion_models/   (or models/unet/)
    │   └── MiniMax-Music3-Q8_0.gguf
    │
    ├── text_encoders/       (or models/clip/)
    │   └── minimax_music3_text_encoder_bf16.safetensors
    │
    └── vae/
        └── minimax_music3_dav.safetensors

🚀 How to Use in ComfyUI

Install the ComfyUI-GGUF extension via ComfyUI Manager.
In your workflow:
- Load the GGUF model using Unet Loader (GGUF).
- Load the text encoder using the standard CLIP / Text Encoder loader.
- Load the VAE using the standard VAE loader.

📥 CLI Download Commands

1. Download GGUF Model

huggingface-cli download Abiray/MiniMax-Music3-GGUF MiniMax-Music3-Q8_0.gguf --local-dir ComfyUI/models/diffusion_models

2. Download Companion Text Encoder & VAE

bash

# Text Encoder
huggingface-cli download Comfy-Org/MiniMax-Music-3 text_encoders/minimax_music3_text_encoder_bf16.safetensors --local-dir ComfyUI/models/text_encoders --local-dir-use-symlinks False

# VAE
huggingface-cli download Comfy-Org/MiniMax-Music-3 vae/minimax_music3_dav.safetensors --local-dir ComfyUI/models/vae --local-dir-use-symlinks False

✍️ Prompting Structure

MiniMax-Music-3 generates the best results when given two separate conditioning inputs:

Lyrics (with section markers on individual lines)

text

[Verse]
Morning light filtering through the pine
Every quiet street is yours and mine

[Chorus]
Softly the world begins to breathe

Music Description

text

Genre: acoustic pop. BPM: 96. Key: C major. Warm and intimate, building gently into the chorus.
Vocals: soft female lead, close and breathy, light stacked harmonies in the chorus.
Arrangement: fingerpicked guitar and soft piano; brushed drums and upright bass enter in the chorus.

📜 Credits & License

Original Model: MiniMaxAI/MiniMax-Music3
ComfyUI Repack: Comfy-Org/MiniMax-Music-3
License: Inherits the Apache-2.0 License from the original release.