MiniMax Releases MiniMax-Music3: An Open-Weights Music Model Generating Complete Five-Minute Songs From Lyr…

What it does

MiniMax launched MiniMax-Music3, a text-to-music AI model that generates complete songs up to five minutes long. It creates 32 kHz, 16-bit stereo WAV files directly from lyrics combined with a structured caption that tags song sections. The open-weights model processes the entire song in a single pass, allowing for coherent long-form audio generation rather than short clips or fragmented outputs. MiniMax-Music3 supports detailed input formatting, giving users control over the song’s structure and narrative flow.

Why it matters

MiniMax-Music3 tackles a persistent challenge in AI music generation—producing long, continuous songs rather than short melodies or loops. By handling up to five minutes of audio in one go, it reduces the need for manual stitching or post-processing, which lowers production overhead for creators and developers. The ability to generate songs from lyrics with section tags adds a layer of semantic control that can enable more meaningful compositions tied to lyrical content. The open-weights release signals a push toward transparency and community experimentation, distinct from black-box proprietary models.

Who it is for

This model targets music creators, researchers, and developers who want to integrate text-driven music generation into apps, games, or experimental projects. It suits operators needing longer, structured songs that align with their narrative or branding needs, without licensing concerns typical of closed commercial offerings. Because MiniMax released the weights openly, it benefits academic research and startups looking to build custom layers or fine-tune the model for specific genres or styles.

The catch

Despite its capabilities, MiniMax-Music3 requires detailed input formatting and technical know-how to leverage its full power. Users must provide lyrics with section tags and companion captions, which adds a workflow step. The open weights are under license conditions that may affect commercial deployment; operators must review them carefully before incorporating the model into products. Finally, the computational cost to generate five minutes of high-quality stereo sound in one pass may be significant, demanding substantial hardware or cloud resources.

What to watch next

Expect rapid exploration on how structured captions and lyrical input can deepen AI music coherence and emotional alignment. Watch for follow-up tools or wrappers that simplify the input formatting process and democratize access for non-technical users. Improvements in runtime efficiency and integration options will be critical for wider adoption. Licensing and community-driven forks will shape the model’s ecosystem, especially around commercial use cases and customization. MiniMax’s approach could also pressure competitors to open more of their models or enable similar long-form generation.

AI Quick Briefs Editorial Desk