MiniMax-Music-3: 320,000 Monthly Downloads: Use MiniMax Music 3 Directly Without Conversion

We provide the original MiniMax Music 3 model files organized in a structure recognized by ComfyUI. Packaged in the Diffusion Single File format, they can be connected directly to ComfyUI workflows without any separate conversion process.

All core components required for the model architecture are included. The DiT model is available in three precisions: fp16, fp32, and int8_convrot, while the text encoder is provided in bf16 and pruned versions. VAE files are also included, addressing all essential elements of the image generation pipeline at once.

Unlike traditional methods, there is no need to download individual files or manually configure folder structures. Simply place the files in the diffusion_models, text_encoders, and vae folders under the ComfyUI models directory. Released under the Apache-2.0 license, it has minimal restrictions for commercial use.

This is suitable for developers building ComfyUI-based image generation environments. It reduces initial setup time when testing MiniMax Music 3 performance or integrating it into production pipelines. It is actively used by the community, recording over 320,000 monthly downloads.