Stability AI launches Stable Audio 3.0 for longer music generation

Stability AI is releasing a new family of audio models called Stability Audio 3.0. According to the company, the top model in this family can generate professional-grade music of more than six minutes long. Both the medium and large models in the release can create full compositions of 6 minutes, 20 seconds long that maintain musical structure and melodic tone. This capability represents a significant increase in length compared to Stable Audio 2.0, which was released in 2024.

The company is releasing four distinct models under the new family:

• Small SFX: 459M parameters, suitable for on-device sound and music generation of up to two minutes.
• Small: 459M parameters, also suitable for on-device sound and music generation of up to two minutes.
• Medium: 1.4B parameters, capable of generating full compositions of 6 minutes, 20 seconds.
• Large: 2.7B parameters, also capable of generating full compositions of 6 minutes, 20 seconds.

Stability AI is making the small SFX, small, and medium models available with open weights—meaning the models are available for public use and modification. This is an upgrade from Stable Audio Open, released in 2024, which allowed for music generation of up to 47 seconds.

In contrast, the large model is available only through the Application Programming Interface (API) and self-hosting paid services. Additionally, companies with more than $1 million in revenue would need to obtain an enterprise license to use the models.

The release comes as a variety of companies, including Google and ElevenLabs, launch models and tools for music generation. However, the long-term survival of these services may depend heavily on data licensing and partnerships with music labels, as highlighted by the ongoing court battles involving competitors Suno and Udio. To address these challenges, Stability AI stated that its latest set of audio models is built on fully licensed data. This follows deals the company signed last year with Warner Music Group and Universal Music Group to develop models and music-creation tools.

The startup is also developing a new suite of products for professional musicians. To lead this professional music offering, Stability AI has hired Ethan Kaplan, the former chief digital officer at Universal Audio and Fender. This hire reflects a broader trend of artificial intelligence companies recruiting music industry executives to strengthen their positions. For example, Suno recently hired Jeremy Sirota, the former CEO of Merlin, as its chief commercial officer, while ElevenLabs hired Derek Cournoyer, formerly of indie music publisher Kobalt, as a strategy lead for its music business.

Why it matters

Stability AI is differentiating its new audio models by emphasizing fully licensed training data, a strategic move to avoid the legal pitfalls currently facing competitors in the generative music space.