MiniMax Music 3 And Full Pipeline With H3 + Music Planner Node - Free Workflows (Patreon)
Downloads
Content
Hey everyone.
New video is up, and this one is for anyone who's ever wanted to make their own music without paying for a subscription. MiniMax just open-sourced Music 3, a model that writes full songs up to five minutes long from lyrics and a style description, and it runs locally in ComfyUI. No Suno, no Udio, no credits. I walk through the whole thing in the video: downloading the model files, the basic text-to-music workflow, and then the full music video pipeline with MiniMax H3.
One thing that surprised me while testing: the lyrics-to-audio timing. If you target 60 seconds in your lyrics, the model will actually cut off the ending. The music intro and the gaps between sections eat up time, so you have to set the max duration longer, around 80 to 90 seconds for a 60-second song. Took me a few tries on the official site before I figured that one out.
I'm sharing two workflows for free right here:
FREE WORKFLOWS (attached below)
WHAT YOU NEED
- MiniMax Music 3 models (diffusion, text encoder, VAE) from the Comfy-Org repo
- MiniMax H3 models (ref2va, audio VAE, video VAE, Qwen3VL text encoder)
- Latest ComfyUI version
- 8GB VRAM minimum for Music 3, quantized int8 versions run on less
MiniMax-Music3
https://modelscope.cn/models/MiniMax/MiniMax-H3
https://huggingface.co/MiniMaxAI/MiniMax-Music3
https://minimax-ai.github.io/music3-demo/
MiniMax-Music3 for ComfyUI
https://huggingface.co/Comfy-Org/MiniMax-Music-3/
ComfyUI MiniMax M3 SongPlanner
https://github.com/benjiyaya/ComfyUI-MiniMax-M3-SongPlanner
Minimax-H3-Prompt-AgentSkill
https://github.com/benjiyaya/Minimax-H3-Prompt-AgentSkill
ComfyUI-H3-VisionPromptor
https://github.com/benjiyaya/ComfyUI-H3-VisionPromptor
TIPS
Set the max duration about 50% longer than your target song length. The model needs room for the intro, the gaps between sections, and the outro. 60 seconds of lyrics needs 80-90 seconds of max duration.
Keep the seed fixed and shared between the text encoder and the KSampler. You get much more consistent output between runs.
The lyrics tags matter. Use [Intro] [Verse] [Chorus] [Bridge] [Outro] to control the song structure. If you want an instrumental, skip the lyrics entirely and just describe the arrangement.
I'll be posting the advanced music video pipeline for Patreon supporters in the member post.
Thanks for watching. See you in the next one.
— Benji