MiniMax-Music3

Note: All model files and custom nodes below must be downloaded via the client. If you do not have the client, Download >>>

MiniMax-Music3 is an open-weight AI music model designed for creating full-length, studio-quality songs from scratch. For non-technical users, its biggest advantage is simplicity and intelligence: by simply providing your lyrics and a brief style prompt (e.g., "upbeat pop with breasthy female vocals and acoustic guitar"), it automatically composes, arranges, performs, and produces a complete track up to five minutes long with structured verses and choruses. It delivers remarkably natural-sounding vocals with lifelike phrasing and breath, while remaining lightweight enough to run smoothly on standard personal computers.

MiniMax-Music-3 vae >1 files

Name	Size	Target Dir
minimax_music3_dav.safetensors	216.7 MB	vae

MiniMax-Music-3 text_encoders >3 files

Name	Size	Target Dir
minimax_music3_text_encoder_pruned_int8_convrot.safetensors	9.2 GB	text_encoders
minimax_music3_text_encoder_pruned_bf16.safetensors	16.71 GB	text_encoders
minimax_music3_text_encoder_bf16.safetensors	18.47 GB	text_encoders

MiniMax-Music-3 >3 files

Name	Size	Target Dir
minimax_music3_dit_int8_convrot.safetensors	2.5 GB	diffusion_models
minimax_music3_dit_fp16.safetensors	4.91 GB	diffusion_models
minimax_music3_dit_fp32.safetensors	9.83 GB	diffusion_models