Stability AI Launches Audio Model Generating Six-Minute Songs

Stability AI has unveiled Stability Audio 3.0, which consists of four audio models that can create high-quality music lasting up to six minutes and 20 seconds. These models vary in size, from small SFX (459M parameters) to large (2.7B parameters), with the smaller versions designed for on-device generation of up to two minutes. The medium and large models can generate compositions that are twice as long as those produced by Stable Audio 2.0 (2024). The small SFX, small, and medium models are open-weight, while the large model is accessible through API and paid services. Enterprises earning over $1 million must obtain an enterprise license. Stability AI has partnered with Warner Music Group and Universal Music Group, utilizing fully licensed data. Ethan Kaplan, who previously served as chief digital officer at Universal Audio and Fender, will lead the professional music initiatives. This launch aligns with industry developments, as Google and ElevenLabs have also introduced music generation tools, while Suno and Udio are embroiled in legal disputes regarding data licensing.