Stability AI has released a new open AI model that can generate sounds and songs. It is called Stable Audio Open. You give it a text description like “Rock beat played in a studio on an acoustic drum kit” and it creates a 47 second recording of that sound.
The model was trained on 486,000 free sound samples from libraries like Freesound and the Free Music Archive. Stability AI says it only used recordings that are royalty-free so there are no copyright issues.
With Stable Audio Open, you can create drum beats, instrument riffs, ambient noises, and other elements for videos and films. You can also use it to edit existing songs by applying a different style to them. However, it cannot generate full songs, melodies, or vocals very well.
The model has some limitations. It works best for descriptions in English and some music styles are represented better than others in the training data. The sounds it creates also cannot be used commercially under the terms of service.
Stability AI is promoting this open-source release, but also trying to get people to pay for their premium Stable Audio product that can do more. The company has faced controversies over the ethics of training AI models on copyrighted works.
About Stability AI: It is a startup that created the popular AI art generator Stable Diffusion. It has struggled financially and is looking to revive its business with new products like this sound generator.
