Unveiling Stability AI’s Model for Generating Seamless Sound: Introducing Stable Audio Open Source Version, Capable of Producing Tracks up to 47 Seconds in Length.

Stability AI has unveiled the Stable Audio Open model, a text-to-audio model in an open-source version trimmed down from the full version of Stable Audio tailored for commercial use.

The key difference between Stable Audio Open and Stable Audio lies in the fact that the Open version can generate sound for 47 seconds compared to the full version’s 3 minutes. Stability AI specifies that the Open model is designed primarily for creating audio samples and sound effects rather than composing full-fledged songs. Examples of the sound can be heard from the source.

The Open model is also trained on data from FreeSound and Free Music Archive, eliminating any copyright issues. The model is now available for use on Hugging Face.

TLDR: Stability AI introduces the Stable Audio Open model, a text-to-audio model tailored for commercial use and capable of generating sound for 47 seconds, trained on data from FreeSound and Free Music Archive.