Music production tool from Stability AI: Stable Audio Open

Stability AI introduced Stable Audio Open in a blog post, saying that the tool is trained entirely on royalty-free music and can take a text description and turn it into a recording up to 47 seconds long. Stable Audio Open is based on the company’s commercial text-to-audio creation tool, Stable Audio, which was first released in September. The company developed the Stable Audio 2.0 tool in April, providing greater clarity to audio outputs and increasing the length of sounds.

Unlike this model, which is trained on copyrighted music sources, Stable Audio Open is only trained on 486 thousand samples from free music libraries such as Free Music Archive and FreeSound, so it is unlikely that the recordings will be of the same quality.

Stability AI said it was designed for musicians and filmmakers who need to create drum beats, instrument riffs, ambient sounds and production elements. The tool can also edit existing songs or apply a new style, such as smooth jazz, to a song of a different style. An interesting advantage of Stable Audio Open is that users can fine-tune the model with their own sounds, meaning a musician can load samples of their own drum recordings or guitar riffs to produce new sounds in their own style.

However, the open nature of Stable Audio Open limits its usefulness to some extent. The company explains that the purpose of the model is not to create full songs, but rather shorter sound bytes that can be used as musical effects. The tool cannot produce a long song or vocal. Those who want to do this would be better off using the premium Stable Audio 2.0 service or an alternative such as the Suno platform, which can produce music longer than three minutes and with more consistency.

Stability AI also admits that the model doesn’t perform very well when people give commands in languages ​​other than English.