Stable Audio

Stable Audio is a music and sound generation product developed by Stability AI using AI music generation models. The product generates audio tracks based on descriptive text prompts from users. Prompts describe the desired audio, including genres, instruments, descriptive phrases, tempos, and track length. The underlying model behind Stable Audio was trained using music and metadata from the library AudioSparx. Via a latent diffusion approach, Stable Audio uses audio conditioned on text metadata as well as file duration and start time to control the content and length of the generated audio. Stable Audio is the first AI music generation product that enables the creation of 44.1 kHz music, and the model creates a unique audio track each time it is used.

Stable Audio was first announced on September 13, 2023, and is available through its own website. The company states the product's mission is to empower creatives by providing tools to aid musical creativity. Upon release, Stable Audio is available in three tiers. With the free basic tier, users can create 20 tracks up to 45 seconds long under a non-commercial license. The professional tier costs $11.99 a month and offers 500 monthly track generations up to 90 seconds. Under the professional tier, users can create music to use in commercial projects, such as incorporating Stable Audio outputs as samples in their own music. A third enterprise plan is also available for organizations that want to negotiate a larger number of track generations of longer length. Enterprise plan customers can also use Stable Audio outputs in commercial projects. Stability AI has also stated plans to release an open-source music generation model trained on different data.