Stability AI enters music with Stable Audio

Stability AI just launched Stable Audio, marking the open-source AI giant’s first official product for music and sound effect generation.

Built on a latent diffusion architecture, the tool generates music and sound effects from simple text descriptions via a straightforward web interface. Users simply type in desired genres, instruments, or vibes, set a target length, and get a rendered track in return.

CD-quality sound with precise timing

Generating audio with AI is notoriously tricky, often resulting in muffled, low-sample-rate output that cuts off awkwardly. Stable Audio takes a different approach by generating 44.1 kHz stereo audio, delivering noticeable clarity compared to earlier experimental models.

To solve the timing problem, Stability AI conditioned its latent diffusion model not just on text prompts, but also on audio duration and start times. This means if you ask for a 95-second post-rock track at 125 BPM with euphoric strings and guitars, it generates audio specifically structured for that exact timeframe.

Access is split into two tiers: a free basic version that lets users create and download tracks up to 45 seconds, and a paid Pro subscription that allows 90-second generations with commercial use rights.

Crucially, Stability AI avoided web scraping controversies by partnering with music library AudioSparx. The model was trained entirely on AudioSparx’s catalog and metadata, with revenue sharing baked into the agreement.

Why it matters

For indie game devs, video creators, and musicians hunting for background tracks or fresh samples, Stable Audio is worth playing with immediately. The barrier to entry is zero, and having a custom 45-second ambient loop generated in seconds is a genuine timesaver.

We think Stability AI played this smartly on the legal front. By partnering directly with AudioSparx, they sidestepped the copyright minefield that continues to plague image and language models, making the tool actually viable for commercial creators from day one.

However, a 90-second ceiling means this is currently a sample generator and audio effect toy box, not a full composition suite. Still, by pinning down duration control and 44.1 kHz quality, Stability AI has set a practical standard for what consumer-facing generative audio should feel like.