GitHub radar
MiniMax Music 3 with all songs! AI vocals!
MiniMaxAI has announced a new open weight model, MiniMax Music 3, which composes full music tracks (up to 5 minutes) including lyrics, chorus, bridge, etc based on input text.
MiniMax Music 3 is an open-weight music composition model developed by MiniMaxAI to compose full songs with musical structure (verse, chorus, bridge, etc.) and vocal. Output is stereo 32KHz 16-bit. It uses an 8B global model for the overall musical structure and a 0.6B local model for detailed sound features. Synthesized with Flow Matching and Flow-VAE methods. Model inference using Diffusers or ComfyUI. Also available as a prebuilt ComfyUI workflow on the Comfy-Org port.
Why a vibe-coder should care
Useful for content creation, it allows you to generate your own music/full song with full structure based on text input. It can be run locally on a computer with NVIDIA GPU (at least 8GB VRAM, 24GB recommended for smooth use) for video, podcast, or other content requiring unique sounds.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up MiniMax Music 3 for me: https://huggingface.co/MiniMaxAI/MiniMax-Music3 — install via Diffusers per the README, generate a test track with vocals from 'upbeat pop song about summer', and show me the result
An NVIDIA GPU with 8 GB of VRAM or more is required; 24 GB or more for comfortable use. A MacBook without a discrete NVIDIA GPU won't work.
Open on Hugging Face▌ More finds