Stability AI Unveils Advanced Audio Models Capable of 6-Minute Compositions

Stability AI has announced the release of its new audio model family, Stability Audio 3.0, which can generate professional-grade music compositions lasting over six minutes. This upgrade marks a significant improvement from its predecessor, Stable Audio 2.0, which was limited to shorter music pieces.

The new lineup consists of four models: two smaller variants with 459 million parameters each, one medium model with 1.4 billion parameters, and a large model featuring 2.7 billion parameters. The smaller models are intended for on-device sound generation, enabling users to create music of up to two minutes. In contrast, both the medium and large models can produce full musical compositions lasting up to 6 minutes and 20 seconds, while maintaining a coherent musical structure and melodic integrity.

Stability AI has made the small SFX, small, and medium models available with open weights, allowing users to modify and utilize them freely. This change follows the release of Stable Audio Open in 2024, which permitted only 47 seconds of music generation. The current offerings represent a substantial upgrade, providing greater creative potential for developers and musicians.

However, access to the large model is restricted to API and paid self-hosting services, requiring enterprise licenses for companies generating over $1 million in revenue. This tiered access reflects Stability AI's strategy to serve both casual users and professional entities.

As competition in music generation technology intensifies, major players like Google and ElevenLabs are launching their own models and tools. However, ongoing legal disputes regarding licensing and partnerships in the music industry present potential challenges for these services. Stability AI's collaborations with Warner Music Group and Universal Music Group, established last year, highlight its commitment to building models on fully licensed data, which may offer a competitive advantage.

To enhance its offerings for professional musicians, Stability AI has appointed Ethan Kaplan, former chief digital officer at Universal Audio and Fender, to lead its new suite of music products. While specific features of this upcoming product line remain undisclosed, Kaplan's experience could significantly influence its direction.

The trend of hiring music industry executives is apparent across various AI companies, exemplified by Suno's recruitment of former Merlin CEO Jeremy Sirota and ElevenLabs' hiring of Derek Cournoyer from Kobalt. This reflects a broader industry acknowledgment of the need for expertise in navigating the complex intersection of technology and music.

As Stability AI continues to innovate in the audio space, the implications for musicians, content creators, and the broader music industry are considerable. The introduction of these models not only expands the possibilities for music generation but also raises questions about the future of music licensing and collaboration in an increasingly digital landscape.