Stable Audio 3.0 Large
Stable Audio 3.0 Large is the hosted member of the third-generation audio family. Its API produces stereo audio at 44.1 kHz and returns a job ID while generation runs.
Model design and access
The API uses polling to obtain the completed audio. A successful submission is separate from a completed result. Keep the job ID with the requested duration and prompt so an application can show the actual generation state.Stable Audio 3.0 Large official model documentation
Stability AI offers Large through its API and enterprise self-hosting agreements. Public downloadable checkpoints are provided for Small and Medium, rather than Large. The release also describes audio editing, continuation and LoRA adaptation across the family.Stability AI: Stable Audio 3.0 release
Practical use
Practical guidance. Large fits music services that want managed generation and can process jobs in the background. A long duration limit is useful for a complete cue, but it does not remove the need to review structure, endings and the fit to a scene. Confirm separate enterprise terms before planning an on-premises deployment.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Provider product |
| Model identifiers | Stable Audio 3 Large |
| Model size | 2.7 billion parameters.Stable Audio 3: official inference repository |
| Maximum sequence | 380 seconds in the official model table.Stable Audio 3: official inference repository |
| Delivery | Asynchronous API; retrieve the result with its generation ID.Stable Audio 3.0 Large official model documentation |
Related models
- Stable Audio 2.5
Stability AI · Stable Audio
A hosted music and sound-generation model in the Stable Audio family.
Launch date not confirmedAPI - Stable Audio 3.0 Medium
Stability AI · Stable Audio
A latent-diffusion model for music and sound generation, editing and continuation.
Launch date not confirmedDownloadable