Stable Audio is Stability AI's family of models that generate music and sound effects from text descriptions, including open-weight versions.
In plain words
Sounds and songs come in different lengths, which many audio generators ignored. Stable Audio uses latent diffusion, like Stable Diffusion does for pictures, and adds timing information so you can ask for a clip of a particular length. The first version could make up to 95 seconds of stereo audio in 8 seconds on one A100 GPU. Stable Audio Open was trained on Creative Commons audio, and the newer Stable Audio 3.0 family uses fully licensed data.
A quick example
Imagine a game developer who needs a 30-second loop of calm forest ambience. They could type a description, set the length to 30 seconds and get a stereo clip ready to test in the game.
Where you'll meet it
- Music and sound effects for games, video and ads.
- Research on open text-to-audio models.
Related
Sources
- PaperFast Timing-Conditioned Latent Audio Diffusion, Evans et al. (arXiv) · read 29 Sept 2026
- PaperStable Audio Open, Evans et al. (arXiv) · read 29 Sept 2026
- DocsStable Audio, Stability AI · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.