Stability AI Unveils Stable Audio 2.5, Its First Enterprise Audio Generation Model

September 10, 2025

Details

  • Stability AI has launched Stable Audio 2.5, its inaugural audio generation model tailored for enterprise-grade sound production and commercial brand needs, on September 10, 2025.
  • The company is collaborating with leading sound branding agency amp, part of Landor Group/WPP, to deliver custom audio to WPP’s extensive client roster, emphasizing unique sonic brand identities for large organizations.
  • The model utilizes Adversarial Relativistic-Contrastive (ARC) training and can generate complex three-minute tracks in less than two seconds on H100 GPUs, offering improved musical structure with clear intros, developments, and outros, as well as enhanced response to specific mood or genre prompts.
  • Building on previous Stable Audio versions, 2.5 introduces features such as audio inpainting, enabling users to upload existing audio files and generate seamless contextual extensions, designed to scale brand audio across numerous customer touchpoints.
  • The service is accessible via StableAudio.com, the Stability AI API, and partner platforms like fal, Replicate, and ComfyUI, with options for enterprise licensing, on-premises deployment, and custom model fine-tuning.

Impact

Stable Audio 2.5 arrives as brands recognize that custom audio increases memorability eightfold, yet adoption remains low at 6% of creative campaigns according to Ipsos. With competitors like Soundverse also targeting enterprise users, Stability AI’s latest model positions the company at the forefront of scalable, secure AI audio solutions for branding and commercial applications. The move suggests an imminent shift toward audio-driven brand differentiation in marketing and related sectors.

Read more