MiniMax Music 3 Hits ComfyUI: Generates 5-Minute Structured Songs with Lyrics
Comfy-Org · reddit · 2026-08-14
MiniMax Music 3 is now live in ComfyUI. By providing lyrics and a sound description, users can generate full 32kHz 16-bit stereo songs up to 5 minutes long, featuring consistent vocal identity and clear section structures (intro/verse/chorus/bridge).
Key Features & Tech Details:
- Structure Control: Supports section tags like [Verse] and [Chorus] to define the song's blueprint.
- Deep Customization: Offers a structured caption format to control global metadata (genre, BPM, emotion), vocal details, and arrangement evolution.
- Under the Hood: Uses a hybrid architecture where an 8B LLM (based on Qwen3-8B) handles long-range song structure, a 0.6B LLM fills in frame-level acoustic details, and a flow-matching + Flow-VAE stage synthesizes audio, preventing structural drift in longer tracks.
Requirements: Requires ComfyUI 0.33.0+ or Comfy Cloud. Model weights are openly available.
Related event: MiniMax Open-Sources Music 3 Model for 5-Minute Song Generation(23 posts)→
More from Multimodal
- Kling 4.0 fully available, shows markedly better motion coherence in fast-action scenes — alifcoder · 2026-10-16
- MJev-Spatial: Single-Pass Classification Beats Autoregressive VLMs on Spatial Reasoning — mindlab-research · 2026-10-16
- H-DDCM: Hadamard Codebooks Make Diffusion-Based Image and Video Compression Scalable — Technion · 2026-10-16
- ComfyUI v0.41.0 ships with Flux3 image model and MiniMax-H3 LightVAE support — Fcking_Chuck · 2026-10-16
- AI-generated short film 'The Dragon Rider' glides cinematically above the clouds — jijo6686 · 2026-10-16
- Realistic candid Midjourney portrait prompt: natural skin, lived-in bedroom light — tisch_eins · 2026-10-16