MiniMax Music 3.0 StudioKeep the generation loop close to the GPU.

A local-first ComfyUI workspace for structured briefs, Japanese vocals, tagged lyrics, and MP3 Library playback.

Studio guide

GitHub README

日本語

✦

Brief → queue → Library

Turn structured production notes and section-tagged lyrics into a real ComfyUI job, then audition the MP3 locally.

⌘

MiniMax-focused

The checked-in workflow keeps the MiniMax Music 3.0 nodes, INT8 model path, and audio output contract visible.

◌

GPU intentional

ComfyUI remains the local worker, so model files and renders stay on the workstation that owns the GPU.

↗

Downstream-ready

The Library MP3 can flow into later SRT and HyperFrames lyric-motion work without making video generation part of the core UI.

A verified local workspace

MiniMax Music 3.0 Studio Compose screen with a Japanese vocal brief

The screenshot is from the running local Studio UI: the Japanese shell is using the Compose surface, MiniMax Music 3.0 is ready through ComfyUI, and the brief is prepared for a Japanese vocal generation.

Read the Studio guide · 日本語ガイド

Keep the boundaries clear

MiniMax model weights stay outside Git. The gateway translates browser requests, ComfyUI owns GPU execution, and the Studio keeps local task metadata plus audio playback in the Library.