Stability AI's Stable Audio 3.0 generates 6-minute songs with open weights

Stability AI, known for Stable Diffusion, has unveiled Stable Audio 3.0, a new family of audio models that significantly expand music generation capabilities. The lineup includes four models: small SFX (459M parameters), small (459M), medium (1.4B), and large (2.7B). The small models can generate up to two minutes of sound effects or music, making them suitable for on-device use. In contrast, the medium and large models can produce full compositions of 6 minutes 20 seconds while maintaining musical structure and melodic tone—more than double the length of Stable Audio 2.0 from 2024. This is a massive improvement over the previous open-source Stable Audio Open, which only offered 47-second clips.

Stability AI is releasing the small SFX, small, and medium models with open weights, allowing anyone to use and modify them freely. The large model, however, is only accessible via API or self-hosted paid services, and companies with over $1 million in revenue require an enterprise license. To address copyright concerns, the company has built these models on fully licensed data, including partnerships with Warner Music Group and Universal Music Group. Additionally, Stability AI hired Ethan Kaplan, former chief digital officer at Universal Audio and Fender, to lead its professional music product suite, signaling a move toward serving professional musicians. This launch positions Stability AI against competitors like Google and ElevenLabs, while navigating the legal challenges faced by Suno and Udio.