Stability AI, developers of Stable Diffusion, have just released Stability Audio 3.0—a series of models that can churn out professional-quality songs lasting over six minutes.

The new family includes four models: small SFX for brief sounds up to two minutes; and medium and large versions capable of crafting six-minute compositions with complex structures and melodies. This marks a significant leap from the previous 47-second limit set by Stable Audio Open in 2024.

While Stability AI is opening up access to some models, the larger ones are only available via API or paid services, potentially excluding smaller companies without deep pockets. Meanwhile, the company has secured deals with major music labels like Warner and Universal to ensure its data is fully licensed.

With this move, AI continues to encroach on human creative territories. But as some of its competitors hire ex-industry bigwigs, it remains to be seen how these collaborations will shape the future of music creation.