Stability AI launches Stable Audio 3.0 for advanced music generation up to six-minute long

Stability AI has released Stable Audio 3.0, a family of open-weight music generation models trained entirely on fully licensed data. This update introduces four distinct models, each built for a specific use case: Stable Audio 3.0 Small SFX for sound effects generation on consumer devices, Stable Audio 3.0 Small for full music composition on-device, Stable Audio 3.0 Medium for higher musicality and longer track creation, and Stable Audio 3.0 Large for high-capacity music generation for platforms needing low latency at scale.

Beyond model diversity, Stable Audio 3.0 Small SFX, Small, and Medium are available as open weights on Hugging Face. Stable Audio 3.0 Large, featuring the most advanced musicality, is accessible via the Stability AI API. All models are governed by the Stability AI Community License, granting users full ownership of their audio outputs, along with freedom to distribute and commercialize them.

Building on a next-generation semantic-acoustic autoencoder, this release supports variable-length audio generation with per-second control, enabling tracks up to six minutes long. For the first time, full on-device music composition is possible offline, eliminating the previous limitation to short samples.