Back to all articles
•Technology

Stability AI Unveils Stability Audio 3.0, Expanding AI Music Generation Capabilities

View original source

Stability AI, the brain behind Stable Diffusion, has released a new suite of audio models named Stability Audio 3.0. This release promises a significant advancement in AI-driven music creation, with models capable of generating music tracks over six minutes long.

  • Four models comprise the Stability Audio 3.0 series: small SFX (459M parameters), small (459M parameters), medium (1.4B parameters), and large (2.7B parameters).
    • Small models: Designed for generating sound and music up to two minutes on-device.
    • Medium and large models: Capable of creating full compositions of six minutes, twenty seconds, surpassing the abilities of previous versions.
  • The open weights for small SFX, small, and medium models allow modification by users, while the large model is restricted to API use or self-hosting with paid services.
  • This release sees Stability AI emphasizing collaborations with music giants, evidenced by its partnerships with Warner Music Group and Universal Music Group for data licensing and development.
  • The company's strategic focus includes a pivot towards tools for professional musicians, under the leadership of Ethan Kaplan, joining Stability AI as the head of its professional music offerings.
  • Amid increasing competition in the AI music generation field, companies like Google and ElevenLabs are also pushing the boundary with new models, while legal complexities around data licensing pose challenges.