MiniMax Music3 Semantic Studio

MiniMax Music3 Semantic Studio

Music3 Semantic Studio is an external ComfyUI custom-node package for MiniMax Music 3 generation design and non-destructive post-generation audio editing.

Semantic Studio is a visual authoring/editor tool, not an automatic prompt generator or one-click song generator. It gives you a graphical workspace for designing, organizing and refining the instructions sent to MiniMax Music 3. You can build the song plan yourself or import Caption / Lyrics prepared with ChatGPT, another LLM or an external editor, then continue editing them visually in Semantic Studio.

Current status:

• Semantic Studio — Timeline / Lyrics / Generation UI implemented
• English / Japanese UI — ComfyUI locale-aware labels implemented; other locales fall back to English
• Audio Editor — unified waveform editor, schema-2 automation, Browser Draft Preview, Edit / Mixer / Effects workspace, selection tools and non-destructive fades implemented
• Built-in DSP — Gain, Compressor, Limiter, EQ / filters, Stereo Width, Reverb and Stereo Delay implemented
• VST3 — optional Windows VST3 host, native plug-in UI and state capture available on demand

Neither Semantic Studio nor the Audio Editor patches ComfyUI core, MiniMax Music3 model code, KSampler, or VAE code.

Installation

cd ComfyUI/custom_nodes
git clone https://github.com/ukr8b3g-cmyk/MiniMax-Music3-Semantic-Studio.git

Restart ComfyUI after install/update. The core node package has no additional mandatory Python runtime dependency. The Windows VST3 host is optional and is installed only when a user explicitly requests it from the VST3 workspace.

Quick start — visual workflow

The normal V1 workflow is:

Import Prompt / design in Semantic Studio
                 ↓
       Queue in Capture mode
                 ↓
          generated AUDIO
                 ↓
       Capture / Freeze Audio
                 ↓
   switch to Frozen to keep the take
                 ↓
          Open Audio Editor
                 ↓
     Edit / Mixer / Effects / VST3
                 ↓
             Save Edits
                 ↓
              Queue
                 ↓
       authoritative edited AUDIO

1. Start from the Semantic Studio node

The compact graph node keeps the main generation controls close to the workflow. Music Seed (AR) controls the MiniMax Music3 autoregressive stage, Seed Behavior selects the normal ComfyUI seed behavior such as Randomize or Fixed, and Duration sets the AR generation ceiling. Import Prompt opens the structured prompt importer; Open Semantic Studio opens the full authoring interface.

The Music3 AR seed is separate from the later KSampler seed. Changing one does not replace the other.

2. Import a prompt, inspect it, then edit Lyrics

Import Prompt is intended for Caption / Lyrics text prepared in another LLM or editor. Paste the material, click Analyze, inspect the detected global settings, vocals and sections in Import Preview, then click **Apply Impor