Stability AI is set to make waves in the music tech sector with the introduction of its Stability Audio 3.0 models, which can create professional-grade music tracks lasting over six minutes. This advancement doubles the composition length achievable with the previous version and enhances the quality and musical structure of the outputs.
The new model family includes four distinct options: small SFX and small models, both with 459 million parameters; a medium model with 1.4 billion parameters; and a large model boasting 2.7 billion parameters. The small models cater to on-device applications, generating sounds suitable for shorter projects, while the medium and large models are designed for more extensive compositions, offering up to six minutes and twenty seconds of continuous music.
In a notable shift, Stability AI is making the smaller models publicly accessible with open weights, allowing users to modify and adapt them as needed. This follows the 2024 release of Stability Audio Open, which had a much shorter output limit of just 47 seconds. The new models signify a leap forward, providing creators with tools that promise greater flexibility and creative potential.
However, access to the large model is restricted to paid API or self-hosting services. Companies with revenues exceeding $1 million must acquire an enterprise license to use this model. This tiered access may shape how smaller developers engage with Stability AI's offerings compared to larger firms.
Stability AI's commitment to responsible data usage is highlighted by its partnerships with Warner Music Group and Universal Music Group, ensuring that the models are built on fully licensed datasets. This strategic move aligns with industry trends, as companies like Google and ElevenLabs also explore music generation tools. Yet, ongoing legal disputes involving services like Suno and Udio underscore the complexities of data licensing and partnerships with music labels that will be crucial for survival in this competitive space.
To strengthen its professional music offerings, Stability AI has appointed Ethan Kaplan, a seasoned executive with experience at Universal Audio and Fender, to lead this initiative. While details on forthcoming products remain scarce, this hiring reflects a broader trend where AI firms are increasingly seeking leadership from the music industry to enhance their credibility and product development.
As the intersection of artificial intelligence and music continues to evolve, Stability AI's new models position the company as a critical player in shaping the future of music creation. These advancements could redefine how artists and producers approach music-making, potentially leading to new creative possibilities in the industry.
