MiniMax Releases MiniMax-Music3: An Open-Weights Music Model Generating Complete Five-Minute Songs From Lyrics and a Structured Caption

MiniMax launched MiniMax-Music3, an open-weights text-to-music model that generates full five-minute songs from lyrics and detailed captions. It produces high-quality stereo WAV outputs and is deployable immediately with available weights and code. The model architecture uses a hybrid large language model and continuous synthesis, targeting industries like gaming, advertising, and music tech.