Stable Audio

[Adoption Assessment] As of mid-2026, community testing and tutorials indicate that Stable Audio 3 has moved from buzz to viable local use, with early proof in LoRA customization and cross-platform deployment; however, long-form audio stability requires more production-level validation. [Practical Use] Generates music and sound effects from text, with controllable lengths (Small ~2 min, Medium up to 6 min 20 s), a dedicated SFX model, audio editing, inpainting, and intuitive LoRA training for personal styles. Outputs are trained on licensed data, fully owned by the user and free for commercial use. [Barriers & Cost] Model weights are open-source (per community and official announcements), running locally on consumer hardware, including MacBook M-series (community demo: 59x real-time on M5 Pro) and even CPU-only setups (via community one-click launcher). Cost is only your own hardware, with no generation caps. Online API pricing has not been publicly disclosed (conservatively, the open model itself is free). [Target Users] Ideal for indie musicians, video editors, game devs, and podcasters who need customizable, license-safe audio and can handle some technical setup.